| kathodyne |
Hi!,
My amp runs in ultralinear (7027a beam tetrodes)
Yesterday i put it in triode mode (g2 via an 100ohm resistor to anode)
The primary of the opt is 6k, the UL winding is about 60 to 70% of this....
Which would be best for my end-tube? because in triode mode it has a different rA.....
- just use the main connection for anode??
or
- use the tapped winding?
Thanks for any comment |
|
|
| Agent327 |
Hi,
I would not recommend using the UL taps.
If your UL tap is 60% then the resulting impedance would be 0.6 x 0.6 x 6K = 2k16, if i am not mistaken.
I am not really sure if this 60 % is turns ratio or impedance ratio...:xeye:
That is a very low value for a triode connected push-pull (i assume) 7027A. The RCA datasheet suggests it would rather see 8K. Your 6k is not so far-off.
The 7027A triode curves indicate Ra is about 2K. For PP these are in series and it is not uncommon in triode amplifiers to choose a 3 or 4 times higher anode load, so 12K would not hurt, distortion-wise. You could use the 4 ohm output tap with an 8 ohm speaker to get close to that.
Did you already listen to it or measure something?
Regards |
|
|
| SY |
6K anode to anode is an excellent load for 7027s in triode. With 450V or so, you should be good for 15 watts.
Just tape up the ends of the UL tap wires. |
|
|
| kathodyne |
thanx for the info
i'll let it run on the anode taps
i'm now running at 380v 62,5ma idle per tube
There is a possibility within the amp to set this bias-current.....
it ran like this when it was still used as beam-tetrode (push-pull)
so could i turn the current up a bit?? (70ma) to get a little more output??
or is this not good for the tubes? because the operating point will shift....
The sound is real nice, better then before |
|
|
| SY |
| Turning up the current will not get you a whit more audio power. It will create more heat. |
|
|
| kathodyne |
okay....
because b+ is the main factor now; it's a triode....
it's hot enough as it is... so i'll leave it like this, it's very gentle now!
will switch back to penthode-mode(beam-tetrode) for power if needed...
thanx SY and others |
|
|
| Johan Potgieter |
Kathodyne,
This is not what you asked, but . . . why go to triode if you were in UL? I do not want to stir up feelings among triode-preferring folks, but if you look at the tube data regarding output and distortion vs. the two modes, you will find that UL has about 85 - 90% of the characteristics of triodes, but quite a higher max. output, which means lower distortion at the same output. This is the beauty of UL.
I can only find my data for KT66s now for comparison, but it will be similar for 7027. In triode mode a a pair will give 13W output at 3% total distortion (Va = 400V), while in UL mode for the same conditions the output will be 28W at 1.8% distortion (GEC data sheets). I have comparative data for KT88 which I cannot find now (is that not always the case?!) but as I can recall the r(a) is 1.8K-ohm for triodes, rising to only 2.5K-ohm for UL (43% taps) while the maximum available output (push-pull) is 40W in UL and only 22W as triodes, or something similar (same H.T. voltage). The point is that, despite the belief of some, one gets very much the same characteristics for UL as for triodes, but with a significantly higher ceiling or, inversely, lower distortion at the same output.
As said, this not to confuse you, but do consider - I do not know what your reasons for change are. I have been using 6L6GCs (equivalent of 7027) now for years (I have been in hi-fi for a half century) in this way with great success, and measurements confirmed the above. If you do change, I concur with the advice previously given by others.
Johan |
|
|
| SY |
| Johan, have you considered sensitivity to load? Speakers do not generally look like nice 8 ohm resistors- they have impedances that vary with frequency and level, not to mention reactance. When you take that into account, the small power loss going from UL to triode is no big deal compared to the reduction in load sensitivity. |
|
|
| rdf |
| quote: | Originally posted by SY
When you take that into account, the small power loss going from UL to triode is no big deal compared to the reduction in load sensitivity. |
Is triode less sensitive to load than UL? I was under the impression local feedback gave UL the advantage here. |
|
|
| cerrem |
Here are some usable numbers to check out...
Using a pair of 7027 tubes in Push-Pull wired in TRIODE....
Using a 6K plate load....
Your "cut-off" voltage is -60...so you "could" bias into Class A by adjusting till yo get to -30 @ 74mA .... WHich is 28W plate dissipation...
Or just stay in Class AB 1 run cooler...
The Power Output is 12.6 Watts ..Clean RMS Watts...at 6K plate load..
Plate resistance AVERAGES over the cycle at roughly 1500 ohms....
If you use the 4 ohm tap into 8 ohm load, as another post suggested, you will get 10W output with a 12K plate load...
Damping Factor at 12K would be about .5 ohm "open loop" ....then you need to calculate the DF after the feedback loop is closed....
Since the output transformer was not really intended for TRIODE operation...Frequency response need to be evaluated..since leakages and inductance now contribute differently to the TRIODE's lower plate resistance... Most likely the high frequency response should be OK...The inductance may be too lw and you may have a soft low frequency...but no way of knowing until evaluating...
Chris |
|
|
| SY |
| quote: | Originally posted by rdf
Is triode less sensitive to load than UL? I was under the impression local feedback gave UL the advantage here. |
Well, if you want to use a local feedback model, the triode (strapped pentode- think of a UL with 100% taps) has more feedback than the 43% UL, right?
There are some very illuminating charts in Crowhurst's "Understanding Hifi Circuits" for the 5881, comparing UL to triodes with variations in load impedance with a reactive component included. As always in engineering, there are tradeoffs between the two modes; triodes have greater max power falloff with load change, but their distortion curves are much flatter at loads straying high. The UL does a bit better with loads straying low. |
|
|
| ray_moth |
| quote: | | In triode mode a a pair will give 13W output at 3% total distortion (Va = 400V), while in UL mode for the same conditions the output will be 28W at 1.8% distortion (GEC data sheets). |
I understand that the way the pentode or UL distortion is made up - its spectrum, if you like - is different from that of triode mode. Triode distortion is mostly 2nd harmonic, which will be mostly eliminated by PP, and 3rd harmonic. Pentode and UL modes also produce 5th, 7th and 9th harmonics, which are unmusical and objectionable. |
|
|
| rdf |
| quote: | Originally posted by SY
Well, if you want to use a local feedback model, the triode (strapped pentode- think of a UL with 100% taps) has more feedback than the 43% UL, right?
|
(Too) obvious once I think about it that way. Thanks for the Crowhurst tip, I'm finding UL resources difficult to come by.
| quote: | Originally posted by ray_moth
...Pentode and UL modes also produce 5th, 7th and 9th harmonics, which are unmusical and objectionable. |
I'ld be interested in seeing measurements. I compared triode to UL on an EL84 using Audiorightmark. Without a scope the peak output levels were unknown but within the 120 dB limits of an Audigy I didn't see this behaviour. There was a noticable increase in 4th or 5th compared to triode but nowhere near what's often suggested. Anyone know of data? |
|
|
| kathodyne |
@Cerrem....
:) wow....:cool: you're right!...high frequency is really nice but i was already wondering why there was a little less bass then before...impedance
so 6k would be a little too low....just as a test i'll listen what it does when using the 8ohm taps with 8 ohm speakers (6-8 ohm they have written on them)
if i understood you people correctly there should be even less sound!
the other sort of distortion ánd because the mid/high is soo nice now i'll be looking around to build a real triode amp.
if just the ship with money came sailing in....;)
For now i'll just install a switch so i'll be able to switch the two operating modes (when the HV is down ofcourse-----wonder what happens but don't dare to try :D )
- don't work on amps with friends nearby ------ they'll find a way to get your soldering iron INTO your circuit- |
|
|
| SY |
| quote: | | I compared triode to UL on an EL84 using Audiorightmark. |
Did you have a load that was reactive, i.e., a reasonable speaker model?
You're right, hard comparisons between UL and other modes are difficult to come by. |
|
|
| rdf |
| No, the load was simple R. Even at that I don't completely trust my results having found significant discrepencies between two of my computers making the same measurement. It's something I want to revisit. |
|
|
| PRR |
> switch the two operating modes (when the HV is down ofcourse
You can switch triode/UL/pentode with power on and signal flowing.
It would be best to use a snap-action non-shorting break-before-make switch. Break the screen wire and the tube goes dead safely.
If you used a shorting make-before-break switch, that was not snap-action: if you linger in the in-between positions, the output transformer is shorted. This does no immediate harm, but if you also have signal flowing and crank it up because nothing is coming out, the tubes will overheat.
It really does want to be a 600V switch. With small tubes and low B+, switching at low-signal moments, a lesser switch will work for a while. But with big signal power the switch will arc enough to burn the contacts of a $1 50V switch after a few operations.
> the way the pentode or UL distortion is made up - its spectrum, if you like - is different from that of triode mode.
Not the mode but the tube. OK, if you swing a pentode right into the low-volt high-amp knee you will find a lot of 7th, 8th and 9th, but that vanishes when you run a few dB lower.
If the grid were perfect, high-order harmonics would be vanishingly small, with or without a screen grid.
Grids are never perfect.
The classic audio triodes like 2A3 and 300B were laid-out with coarse grid pitch so that minor variations in pitch due to grid lathe wobble would cause only minor kinks.
The "hot" pentodes were not intended to be used for "good" audio without feedback. With feedback, or for "bad" audio (cheap radios), what really matters is gain. They use a close-pitch grid that gives more gain, but is also more sensitive to grid-lathe wobble.
I'm thinking of 1930s production. 1950s grid lathes were good and old 6L6 are pretty linear for a tube that was not made to be super-linear, though some of the low-price tubes like 50L6 may have been assembled on worn-out lathes to keep costs down. 1990s production is small-market low-tech, and some grid lathes still in use may not even be good enough for a 2A3 to show really low high-order harmonics.
> wondering why there was a little less bass then before...impedance
Still wondering. The triode's low Rp should give better bass, because it is not loaded by OPT inductance so much. However hi-fi p-p OPTs usually have way more inductance than they "need", to keep distortion down. Maybe what you are hearing is not less bass, but less bass distortion? That could happen if the OPT is not oversized enough. |
|
|
| Johan Potgieter |
| quote: | Originally posted by SY
Johan, have you considered sensitivity to load? Speakers do not generally look like nice 8 ohm resistors- they have impedances that vary with frequency and level, not to mention reactance. When you take that into account, the small power loss going from UL to triode is no big deal compared to the reduction in load sensitivity. |
You are quite right about the atrocious loads that loudspeakers present. But: The power output for UL could be a little higher than for straight pentodes because the screen is now also contributing somewhat to output power, while the triode output under similar conditions would only be about 30% of the above. At about 35% taps the r(a) [plate resistance; I cannot do subscripts] on the other hand would only be about 10% - 15% higher than that for a triode. Since that is what makes a stage "load insensitive", you have almost the same load tolerance as for triodes, but with a distinct nett advantage in sensitivity and more important, output power and distortion. The power difference is not so small, while the other advantages exist.
I am so sorry that I cannot like a decent chap back this up now with figures; I will really hunt for a very illustrative GEC graphs for KT88s that I have here somewhere and give figures as soon as possible. Meanwhile you will have to accept the figures quoted from memory in my earlier post. To sum up, with UL (and proper design!) one has most of the advantages of pentodes (tetrodes) and most of the advantages of triodes. That is because, as the screen is moved from B+ towards the anode, the available output power decreases slowly (after even an initial increase as explained above) and only more rapidly after a 50% tap, while the r(a) deceases rapidly and flattens out later.
To venture a further observation: Certain amplifiers give a switchable option between triodes and UL. As far as I know only the screens are switched between UL taps and anodes and folks witness to hearing the difference (in favour of triodes). If such switching is true, this is an over-simplified comparison. Neither the load nor the current is the same for optimal conditions in either mode.
I would say again: I do not disparage the use of triodes (have done so myself), but I truely cannot see myself preferring those in favour of UL. The type of tube also plays a role, but only slightly so. But let me see if I can come up with figures (might even be able to display the graphs here).
Johan |
|
|
| Johan Potgieter |
I have read post #17 only after I posted (#18) earlier.
Good post by PRR.
Just to remember that pentodes were initially made not primarily for low distortion but high frequency (RF) (small pentodes) with realisable advantages for audio (bigger pentodes). High fidelity as we know it only came later, I think with the Williamson. The beam tube principle (a little more linear it would appear) was also later, although (way back!) I think the 6L6 was already there before 1940.
Yes, UL graphs are difficult to come by (I mean distortion-wise). How I wish I had the spectrum analysers that my work could afford! I am at present busy with a 100W amp using 4 x 6L6GC. Perhaps I can endear myself at the local university and score a few hours on their equipment.
Johan |
|
|
| PRR |
> High fidelity as we know it only came later, I think with the Williamson.
That's when it was called high fidelity. But the 2nd-generation movie sound systems, early 1930s, were very faithful reproducers, within their design goals (dispersion is "bad" when you have limited power and over-live acoustics). Fantasia was played on these systems with stunning effect. These were often hot push-pull triodes. The rise of Pentodes was the decline of audio, what the hi-fi nuts of the 1950s were fighting against.
> pentodes were initially made not primarily for low distortion but high frequency (RF) (small pentodes) with realisable advantages for audio (bigger pentodes).
Pentodes happened because triodes and tuned circuits don't play well together. And not just because triode Rp damps the tank. If the input and output tanks are not right on the same frequency (they never are), you get over 45 degree phase-shift in each, which with the plate-grid feedback (1/Mu) and the tube's inverting gain gives everything needed for oscillation.
A second RF-grounded grid kills the feedback path. If left at DC ground, it also kills current, because the grid-cathode interface feels the extra grid voltage more than the plate voltage, and electrons have no reason to wander across. If taken to or past plate voltage, it sucks all the electrons. But kept at a high voltage, lower than plate, with AC grounding, you get a good tube that can give high RF gain without instability.
In audio, the tetrode separates two functions: urging the electrons away from the cathode, and collecting electrons to heat-up the load. Plate voltage can drop while current is rising. Gain can be high.
Problem is that the screen has to sit at a high voltage and the plate has to stay above that voltage or the screen will capture ALL the electrons. Another problem is that electrons that hit the plate knock secondary electrons loose and flood the screen.
The fix for that new problem was a third grid, near DC ground. Screen (G2) is run positive enough to pull electrons away from the cathode and throw them through the negative region to the plate. Secondary electrons are low-energy and the negative region turns them back to the plate. G3 also adds another layer of shielding for incredibly low feedback. However, G2 current tends to be a significant fraction of total current, reducing Gm a little, complicating wiring, eating power, and causing huge extra noise.
> The beam tube principle (a little more linear it would appear) was also later, although (way back!) I think the 6L6 was already there before 1940.
Beam "tetrode" tube is just a different way to build a pentode, apparently to avoid certain patents on 3-grid power tubes. Instead of a third grid near DC ground, they get essentially the same effect by forcing the electrons into a thin beam. The electron's own charge gives a sufficiently negative area to act as a G3, without looking anything like the patent drawings for power pentodes. At a high level, you could say that RCA just replaced the several-turn horizontal electrode with a 2-element vertical electrode structure, and the electrons don't know the difference. To cover their rears, RCA found several other small improvements and wrapped them up (around 1937) as the 807/6L6, and provided extensive write-ups to show its novelty. In many ways (not so much for fidelity) it was a very superior power tube. But if you compare the early 807/6L6 curves to Euro Power Pentodes and then to the later 6L6GC, you suspect that RCA finally abandoned the extreme beam-forming proportions, and OTOH the Euro tube makers adopted the 6L6's aligned-grid trick.
> I wish I had the spectrum analysers
A PC with a sound card can do spectrum analysis. I was doing it on 386s, though a hot 486 is a much better engine. There are several packages, some freeware. One (RightMark?) is primarily designed to test soundcards with a loopback cable, but you can put anything you want "in the cable" if you can get the overall gain to unity. Working with Power Amps, you need some care to attenuate the signal so you don't smell smoke from the PC. You are of course limited by sound card performance, but I've had $19 no-name USB sound-boxes show THD below 0.05%, which is good enough for most power tube work. (I got the $19 USB-sound because an on-board audio claimed 24-bit but actually delivered 15-bit, as if the top bit of the ADC was stuck.) |
|
|
| Johan Potgieter |
Quite so, PRR.
We are a little off-thread, but hopefully still interesting! I personally fancy the beam tube style because of the slightly higher efficiency, but mainly because I experienced with especially EL34s, that the g2 would occasionally glow in certain places where it was "unprotected" by a screening g1 winding. This often resulted in spurious oscillatory response, so I could not see any reason to stick to classic pentodes. There did not appear to be any "alignment" as such with G1, and in some tubes the relative placing would allow for several consecutive screen turns to be open to the cathode. (I am always collecting failed tubes and cutting them open.)
But back to UL vs triodes, if Kathodyne has not yet thrown up his hands in despair. I will be posting the promised figures shortly. |
|
|
| Miles Prower |
some of the low-price tubes like 50L6 may have been assembled on worn-out lathes to keep costs down.
That's not the problem. The 50L6 (and its 7-pin miniature cousin, the 50C5) were designed for use with the "suicide box" topology of directly rectifying the AC main without a power xfmr. That means low voltage, and the 50L6 (50C5) are rated at 110V(DC). It can get out a few watts, but the linearity is horrible.
I've seen some designs, done for S's 'n' G's, that attempt to improve the audio performance. What these do is bust the specs, and run them at 250V(DC), with the plate current reduced to keep the dissipation within ratings. Furthermore, these designs tend to resemble a solid state design: preamp -> VAS -> 50[CL][56], and use way more feedback than is normal for a VT design (40db+) to linearize those beasts. Still, the performance leaves a lot to be desired. Often offered as low power guitar amps since it's necessary to limit the bandwidth to prevent the OPT's from going funky with all that NFB. |
|
|
| Johan Potgieter |
Miles?
40 dB of feedback????!! And I thought I have seen it all in my little sphere of audio. How on earth did they get that round a (I suppose not the best quality) output transformer? I thought going past 26 dB was pushing it.
Johan |
|
|
| Miles Prower |
How on earth did they get that round a (I suppose not the best quality) output transformer? I thought going past 26 dB was pushing it.
OK, I found the original web site, Here. As you can see from the schematic, there is one helluvalot of open loop gain there. Considering that a 50C5 (and cousins) typically work at a V(gk)= -7.5V (15Vp-p audio) you don't need that much open loop gain unless you're really going to pour on the NFB, which is what he does.
As for how to get that around an OPT, he limits the bandwidth at both ends so as to avoid the large phase shifts that occur at frequency extremes. This, in turn, makes the thing useless for almost everything but a guitar amp.
As I said before, it's mainly a "let's see if I can do it" type project. There are way better tubes for an amp of that power level (6AQ5A -- down-rated 6V6 essentially -- would be far better).
Schematic:
 |
|
|
| Johan Potgieter |
Thanks Miles,
Oh Boy . . . . . .
(In case of possible misunderstanding: I did not doubt your word, Miles, just expressing incredulity.) |
|
|
| Miles Prower |
It is pretty incredible. A whole lot of bother that could be avoided simply by going with some decent tubes in the first place. Oh well, I guess it's OK if you have a lot of free time, and want the braggin' rights.
Wouldn't be my choice. :whazzat: |
|
|
| kathodyne |
Hello,
Monday i went to Mattijs de Vries (www.machmat.com), he did some measurements on my amp.....:wiz:
bandwidth -3b: @10hz-@80khz
output (triode PP): approx 12w @62,5ma 380v
distortion: 0,68% @70% output
-distortion is exactly symmetrical- many thanx to the longtailed pair splitter with current source (5965 @100v 5ma) :wave2:
Some changes have been made to the amp:
OLD PSU = 2xpy500A - 47uf - 10H - 300uf - 1k5 - 82uf
NEW PSU= 2xpy500A - 220nF - 47uf -10H - 450uf - 220nf - 3k3 - 82uf - 220nf
the increase of capacitance for the output stage with the adding of little fast ones solved the bass issue...got plenty now!
the grid resistors of the 7027a from metalfilm to allen&bradley's
the current has been turned op to 70ma
and the coupling capacitors have been changed from auricap to PIO
Sorry for eventually subjective use of language; i find it very difficult to translate what i hear into words....here goes
these changes together make the amp sound really nice!! i'm listening to my music collection non-stop!! the PIO condensers were a little bit cloudy at first and things got a little unclear (although high freq backgroundsounds in music were like 3d immediately!!!) now after 2 days they're beginning to show their potency...it's gettin better!!
And a question....why feedback?? it's an error solving device...you send your signal back into the parts line, thus getting more 'parts sound'; stability problems; loss of sound...try not to make an error....then you don't have to correct it.
I'd like to see what you people think about feedback
-@prr@cerrem@johan potgieter : i very much appreciate your answers and the loads of information that i can look into because of them....
Unfortunately i just started with tubes about 2 years ago; with no technical/electrical study whatsoever...
Now i can calculate a common cathode -simple- schematic and i know what is going on in my amp and why things happen.
this is because i read and i read and i read and measure my amp and try to calculate these measurements...so i'm collecting bits of information and try to link these together with common sense and formulas i find...
the point? i try to understand but am not able to give an answer that makes sense
So, i wonder are there people who'd like to device some tube-setups which can be built at home by beginners....then they answer the questions given...which could be corrected and faults in thinking about the tubes would come above...a bit like a tube course?? in addition to the info for newbies (say moderates :)) |
|
|
| Miles Prower |
And a question....why feedback?? it's an error solving device...you send your signal back into the parts line, thus getting more 'parts sound'; stability problems; loss of sound...try not to make an error....then you don't have to correct it.
I'd like to see what you people think about feedback
Yes, feedback does correct for errors, amoung other things. Additional benefits include noise reduction, increased bandwidth, and reduced output impedance -- all important. However, it is not a panacea for all ills.
The A Number One problem with solid state is that transistors have incredible gains. Small signal triodes max out at amplification factors of about 100 (6AV6, 12AX7A, 6SF5). High gain pentodes max out at around u= 1100 (12BY7A, 6AG7). A common BJT, even run at emitter currents as low as 1.0mA, will have u= 3076. Run that current up to 50mA, and u= 153846. Having all that gain available is a great temptation to just slap a SS circuit together, and if the numbers come out looking awful, just pour on the feedback. (In the 1950s this was also a problem, though to a lesser extent, with VT designs as well.) Force the numbers to add up, but the sonics are uniformly terrible.
However, judicious use of feedback, after first making sure that the open loop operation has been made as good as practical, can work wonders. This corrects for inherent non-linearities that always exist. Tubes are good, but not perfect. It can also help linearize the otherwise non-linear behaviour of output matching xfmrs. For designs that use pentodes, feedback can certainly help with speaker damping by reducing impedance. (Corrects for exaggerated bass response)
So, i wonder are there people who'd like to device some tube-setups which can be built at home by beginners....then they answer the questions given...which could be corrected and faults in thinking about the tubes would come above...a bit like a tube course??
Not so easy as with transistors, which are real easy to prototype with on a solderless prototyping board. Furthermore, I don't think it's such a good idea given the voltages involved. Every "bread board" set-up I've seen on the 'Net looks like an accident waiting to happen. :hot:
Much better to build the circuit in place with tack soldering. Then you can do corrections/optimizations, and finally complete the circuit once you have it operating satisfactorily. That way, you won't be left wondering where the high voltage is. It's ceratinly safer than the rat's nests of wiring I often see with VT bread boards. |
|
|
| Johan Potgieter |
On Feedback
No need to feel "inadequate". Let me just say that I have been in design for 50 years (not bragging, lamenting!) and it would sure seem natural that one has picked up a few things along the way. Your turn will come.
There are whole chapters on this, thus a little difficult to explain in this space! Perhaps just a few warnings: Negative feedback is the most powerful tool available to improve an amp, and thus also the most abused. Be careful of some of the junk you will find expressed in this regard, especially on the internet. (I am referring to the web in general, not specifically to the DIY Audio site.) There is excellent advice, but also nonsense, and it may be difficult to notice the difference.
No amplifier is absolutely linear, and (negative) feedback simply linearises the design, thus decreasing distortion. Because it is self-controlling, it can never be perfect. There is the "feedback is bad" brigade. Feedback is only bad when too much is used with circuits that are inferior to begin with. But there can be stability problems at either end of the audio band, where the fed-back signal can eventually go in-phase because of phase shifts caused by mainly capacitors (also the output transformer in tube circuits). This will result in oscillation, or at least in spurious instability under certain signal conditions.
It is the job of the designer to see that response at band-ends is such that stability is always assured, and there are simple rules to achieve this. The basics are: The circuit should, before feedback, have negligible phase shift over the audio band. Then feedback is applied and one can perform the necessary tricks outside the audio band to keep things in order. This is relatively simple, and proven principles have been there since the beginning of electronic amplification. The distortion generated in any proper design should be low enough before feedback, that no more than 26 - 30 dB is needed to render an entirely acceptable product (that is a voltage factor of 20 - 30). I repeat that this is entirely feasible; it has been done in a multitude of successful designs, both tube and semiconductor, over many decades.
You mention putting the signal through the same imperfect components several times: Don't be worried about that. It is a very pious idea (one of many floating around) that since no component is perfect, every component must deteriorate the signal. Nonsense. Our ears are extremely sensitive to certain artifacts, but are mercifully still limited. The "imperfections" caused by good quality modern resistors and capacitors are orders of magnitude below this. E.g. the "extra" deterioration caused by the few extra components used to implement feedback may be say 0.1%, while the gain in quality is the feedback factor, say 20x. So there.
There are a few successful tube designs (I must watch the length of this) which I can pass on (perhaps off-site); practical construction will depend on available components, especially the output transformer. Semiconductor designs are freely available, e.g. the Douglas Self designs, but also others.
As said this topic cannot be covered here up to design level, but I trust these few basic remarks helped. I am certain there are other honourable members on-site that can comment. |
|
|
| ray_moth |
| Excellent post, Johan! There is indeed a lot of nonsense spoken and printed about the 'evils' of NFB. It's important to keep things in perspective. |
|
|
| kathodyne |
It seems i've been talking with the 'feedback is bad brigade' !!!!
So, every tube and transistor will have non-linearity...gain of quality is much bigger than the loss by injecting it through the components again (but it will encounter the non-linearity of the tube again.......)
is feedback always in the region of 20-30db??? because that would mean you'll get 6-9 times less output (-3db is halving the power)
So it seems the trick would be to build an amp that is very stable and lineair of itsself and just implement as little feedback as possible.....
(i don't have a clue how to use nfb, so it will be without for now)
to get the most linearity use a current-source on the output and input tubes (as this has the effect of horizontalizing the load line)??? and then one could implement if necessary a little feedback....
@johan potgieter,@miles prower: i was not thinking of a bread-bord....the amps shall be built on wood and with point to point soldering (like always with a test-amp).
it doen't have to be a complete amp, it can be just one stage...
I am still very interested in some 'test'- questions (off-site)
I really want to grasp this...and am very worried that i will teach myself 'wrong' things
btw.: I Respect voltages!! and always work with one hand!! :)
@johan potgieter: I've seen that outpout xfrmr'can cause oscillation; the original xfrmr in the schematic i built caused this...it has been changed from 10k to 6k and....no more problems! |
|
|
| Johan Potgieter |
A few more thoughts . . .
A little more technical: Worst-case distortion occurs when the device internal resistance (Rp in tubes) equals the load resistor. Thus, typically with pentodes or transistors where the Rp is high one uses a load resistor significantly lower in value. With low mu triodes (there is no transistor equivalent) one uses a relatively high load resistor (but there are also other factors governing these values). Linearity will be improved with current sources for loads, but again one cannot simply do that and then create a low output impedance with feedback where such is required. Low impedances created with lots of feedback have their own pitfalls - nothing in this game is ever simple! Doing the above in an output stage to get a high damping factor can lead to all sorts of nasties when the load is a loudspeaker. Prof Matti Otala wrote a landmark article on this (the amplifier-loudspeaker interface) some 25 years ago, but again this will become long and boring for some to fully explain here.
Feedback: Don't see this as passing the signal around and around. In its simplest form, where the gain is high the fed back signal is high, where the gain is low the fed back signal is low. Thus the result with feedback is more linear. The fed back signal can never completely cancel the input, then there will be no output! Thus the fed back signal cancels at most part of the problem, although it can be a large part.
Feedback is not always 20 - 26 dB. Some semiconductor amps use 60 - 70 dB! But to me that is an indication that something is wrong; why is so much needed? The above figure mainly comes from the fact that a well designed amplifier will have distortion of the order of 1 - 2% before global feedback. If we consider that distortion of 0.05% is inaudible, that gives a factor of 20 (26 dB). But I must immediately add that this is an over-simplified statement. The nature of the distortion (high-order harmonics or not plus many other factors) plays a large role.
I must also not generate a false impression re output transformers. When a specific transformer causes oscillations in a certain circuit, it does not necessarily mean that the transformer is inferior. It simply means that the loop phase shifts must be analysed and stability restored by the proper choice of phase correcting components. E.g. when a wrong value capacitor causes oscillations, it does not mean that there is a fault with the capacitor, only with the value.
I am by no means the only one (or the best) here with knowledge, but should you desire you may contact me direct (off-site) with questions, where matters can be explained without the limitation of space. |
|
|
| PRR |
> A common BJT, even run at emitter currents as low as 1.0mA, will have u= 3076. Run that current up to 50mA, and u= 153846.
I think you will find that Early effect limits BJT Mu to 300-3,000, at any current. Certainly not 150K.
Of course for another $0.02, you can cascode to Mu of a million.
> It is the job of the designer to see ...that stability is always assured, ... proven principles have been there since the beginning of electronic amplification.
Just to be pedantic: we had electronic amplification and even some use of feedback before Black and others laid out the theory of why it works and what criteria control stability. There's maybe a 10 year gap between wide use of electronic amplifiers and wide use of feedback theory.
OTOH, you could argue that feedback principles were discovered long before electrons. The speed governor on a steam engine is a feedback system, and sure can become unstable. But simple low-gain systems can usually be made stable with simple hacks; tubes and near-perfect electric reactances made fancy instability possible and easy, and easier to analyze than practical mechanical systems.
> So it seems the trick would be to..... ...... ...... ......
There is no trick! Just different ways to skin the cat. Fashion plays a large part. Whatever you do today will be laughed-at in 20 years, and may be revered in 40 years. Who remembers 1970, the race to point zero-zero-zero THD? Who knew that 20 years later naked SETs with 10%THD would be cool again? Losing as many coupling transformers as possible was one of the early "optimizations", yet iron is in again.
Do what you feel comfortable with.
> blast from the past
I respect Crowhurst and feel everybody should read about everything he wrote, he too was blown around by the winds of fashion. Less than others, because he pretty much knew what he wrote about, but take it as the view of one man at one time.
> just implement as little feedback as possible.....
My current fashion sense says that "small feedback" is bad. Either use NONE, or use a LOT.
Injecting a little output back to the input does raise the issue of re-re-distorting the signal. IM distortion on a simple one-stage amplifier actually rises for 6dB NFB. It may not get back down to no-NFB levels until you get to 30dB-40dB NFB. I think that is an over-simplification, and I know I have enjoyed many low-NFB amps. It is pretty inevitable using pentodes on modern speakers: you need NFB for damping, but you can't get huge NFB around an output transformer.
> Worst-case distortion occurs when the device internal resistance (Rp in tubes) equals the load resistor.
I don't see it that way.
For triodes, THD is caused by current swing, and reduced by plate voltage swing. In small-signal work, you can just make signal current swing small compared to idle current. In large signal work, economics force you to a large signal current swing. Then THD is reduced by large plate voltage swing, i.e. hi-Z loading, and is worst for low-Z loads.
For the other devices, "Rp" is meaningless in power stages. (It may set the no-load gain and this one corner of the stability envelope for a practical amplifier, but becomes irrelevant under load.)
For pentodes, for medium power, percent current swing determines THD, and plate voltage swing has no THD reduction effect. For maximum power, you cut down into the knee of the pentode curve, which has a reverse curvature to the main current-swing nonlinearity. The reverse curve part-cancels the 2nd but raises the 3rd. For best power at lowest THD number, you pick a point in the knee that gives best THD cancellation. This is a paper exercise: THD will vary with minor change of load shifting you to either side of the knee. Also lowest THD is not best-sound: the 2nd is inoffensive, the 3rd is edgy.
For BJTs and FETs, the knee is so low-volt and so sharp that no THD cancellation is possible. Particularly for BJTs, the distortion due to current swing is so bad that we never run them without feedback, at least local feedback, which becomes a whole different problem. |
|
|
| SY |
| quote: | | OTOH, you could argue that feedback principles were discovered long before electrons. The speed governor on a steam engine is a feedback system, and sure can become unstable. |
I would guess that you're familiar with Maxwell's papers on this. Interesting how much of Black he anticipated... |
|
|
| Miles Prower |
Who remembers 1970, the race to point zero-zero-zero THD? Who knew that 20 years later naked SETs with 10%THD would be cool again? Losing as many coupling transformers as possible was one of the early "optimizations", yet iron is in again.
The one thing is probably related to the other. That monomaniacal obsession with ultra-low THD numbers (blame the marketing weenies for that) led to a generation of bad sound. Folks decided to try something different. As for "iron is in again", today's iron is better than ever. Lundahl/MagnaQuest/OneElectron/Sowter/etc. all have access to exotic materials that didn't exist back when they were trying to eliminate as much "iron" as possible.
I respect Crowhurst and feel everybody should read about everything he wrote, he too was blown around by the winds of fashion. Less than others, because he pretty much knew what he wrote about, but take it as the view of one man at one time.
Suspect the demands of the "bottom line" here, more so than "fashion". Good practices remain good practices regardless of what absurd ideas the public-at-large falls for. At one time, everyone wanted SW receivers that came loaded with enormous amounts of gain ahead of the first detector. That remains a bad practice, and you don't need any front end gain (and even then, just a few db's) above 40m (and then only if you are blessed with an extraordinarily quiet QTH). You don't have to fall for anything: if it doesn't improve performance, then don't do it. If it helps, then go with that.
That's the advantage of DiY. |
|
|
| Johan Potgieter |
The dangers of efforts to be brief and not waste readers' time with with extensive cutting/pasting of previous narrative. . . . . .
PRR: I don't feel that we DISagree, rather that you went W - A - Y further than I did. I tried to reply very basically to requests from Kathodyne (posts #27/28) as a starting point - only! I believe one should first get the basics right before advancing to specific applications, and I got the impression that that was what he wanted at the time.
You said that you do not see the matter of R.int/R.load the same as I do, but in your (valid) expansion, for a triode high voltage/small current for least distortion conformed to what I said, and for pentodes the opposite - basically! I commented along tube lines since that was the subject initially. Anyway, let us not enter into a sparring match over this - not my desire. There are obviously specific ways of dealing with every case, after one has the basics right. We also agree, I think that THD is useless especially with semiconductor circuits; the various harmonic products present are what counts.
To move on, the "development" over the years astounds. Single stage output triodes were superceded by push-pull because of obvious sonic advantages, but now we are back with SET - exactly the same thing as 70 years ago! (We are now rather in the thread "The many faces of distortion", but these things are never watertight). I think it is recognised that the reason is that a small quantity of 2nd and 3rd "additions" actually add to the so-called sweetness/musicality sound of this branch - not all distortion is objectionable! . The signal is no longer an exact replica of the original - but it is the customer's money, which he can spend any way he pleases.
But a few years ago HiFi and RV published a list of the 5 best equipment reviewed by some 5 of their top critics, plus their personal choice. One may presume that those gentlemen knew something about realism and faithful reproduction. Yet the top preferences of some did not even appear on the list of others! As a designer, for whose taste must I design? One gets nowhere with this; one can merely design a "clean" amplifier and hope the customer is satisfied. Perhaps I was singularly lucky in that so far all my customers were highly appreciative. Individuals can of course tinker and adapt to their taste.
About what happens at band ends (specifically the high end); this is again a quite complex story, but many of the compromising results can be attributed to what happens to super-audio signals. They operate in a part of the frequency region above say 20 KHz, where they can cause severe (audible) difference products, as has already been alluded to here. I personally believe that this is the reason for much of listener fatigue, with CDs capable of containing a whole band of signals up there. For this reason I am a staunch supporter of proper filtering (cut-off) to cut anything above 20 KHz. Blind tests did reveal that greater satisfaction was obtained with such filters in place than without them. It will absolutely depend on amplifier design, but I am rather safe than sorry. |
|
|
| PRR |
> let us not enter into a sparring match over this
Oh, foo, I didn't mean to be so in-yer-face.
But if I make an objection, I "need" to be sure of it, and sometimes I think out loud at excessive length.
But I still must object to "Worst-case distortion occurs when the device internal resistance (Rp in tubes) equals the load resistor." It implies that low distortion operation is possible in low-Z, lower than when Rl=Rp.
In-context (apparently push-pull), I don't object much. I must have had SET on the mind. In push-pull, Rl matters much less except as it kills power.
When you get to Pentodes(*), Rp in a loaded power-amp is just meaningless. No useful Rl is anywhere near Rp, nor is there any good ratio of Rl/Rp. (At least for basic design; the way Rp falls around the knee in some EL-series tubes is interesting to guitarists who live around the knee.) (* "Pentode" here includes BJTs and FETs, though UltraLinear is in-between.)
> not all distortion is objectionable!
Not even audible. Yet other distortion grates even at very low level.
> Single stage output triodes were superceded by push-pull because of obvious sonic advantages
I'm not sure of that. No hard evidence. I wasn't actually there. But I think P-P soon dominated "high power" work because it was cheaper, and because it measured "lower THD". In short: To Save Money. But is 2% 3rd harmonic less offensive than 5% 2nd harmonic? Depends a LOT on the music, and a lot on your experiences (I've had phases where I loved the 3rd on the right music). But my guess is that 5% 2nd "should be" less-nasty than 2% 3rd, as a general rule.
> As a designer, for whose taste must I design?
Not being personal, but in general: Assuming you design for profit, you have two options. Maximum buzz-words (high selling price), or maximum value for money (high volume). While "clean" is a generally safe goal, I'm not even sure that is what all buyers want. Or rather: a bit of an "edge" will sometimes stand-out as "better", at least to somebody. While I like to theorize the SET fad is about low-low high harmonics and low musical-IMD, I do suspect the sweet 2nd going into 3rd at 2 Watts is part of the attraction for people who grew up on "clean" amplifiers (unlike the 50L6 of my youth). |
|
|
| beamnet |
Incredible thread!
I'm learning a great deal
One of my early teachers ("triode dick") is a non-feedback guy.
i built his kt88 design
Maybe of some value to this thread; i switched from UL to triode to. actually a strapped pentode (?)
I know little about the details, just made the bold move to swap the kt88 for GU50's.
I've been listening to the amp for a few weeks now. I think it sounds good. But then again, it's my first amp and don't know houw good it "should" sound.
Th bass is very neat. No hummy humms and lots of "pop".
But i think it might be a little weak, as my highs sometimes get a little metallic.
I can think of a few scenario's:
- I completely messed up with the GU50's
- The GU50's need bigger coupling caps
- My speakers suck :)
-I'll have to get used to "real" sound and not the exaggerated bass of commercial amps
I'll post the schematic. Maybe somebody could criticise in a way that makes the output stage better understandible for me and everybody else that's relatively new to this?
Bas |
|
|
| kathodyne |
i rebuilt his annastasia... http://kathodyne.coconia.net
(didn't update lately but quite some pictures)
But about the schematic, they don't appear to be in triode.....then g2 should be connected to b+(anode) (and if necessary a 100ohm resistor between it to prevent oscillation)
Did you look at the bias for the gu50?? or are they directly replaceable?
you can measure if the bandwidth is correct by placing 8 or 4 ohm resistor at the speaker outputs (big resistors)
then let it give about 3v at the outputs....connect a scope at the output and a tone generator at the input and sweep through the frequency to see when it starts falling on both sides of the spectrum
you can also calculate the bandwidth of the coupling capacitors (but i don't remember the formula now :xeye: )
something with 1/(2xpixrxc)
you can also look at distortion then by cranking up volume...at clipping both ends of the sinuses have to be flattened equally (which i doubt this splitter-thing will do) |
|
|
| beamnet |
You are right, they are in pentode, not triode, my mistake.
"this splitter thing" is a well balanced long tailed pair. Not an uncommon splitter.
The bias is adjustable with NRS and st to 62 mA or so
The size of the coupling cap depends on the output impendance of the splitter and the input impendance of the output tubes, but i can't find the correct formula's anymore, nor can i calculate the impendances. |
|
|
| PRR |
> "this splitter thing" is a well balanced long tailed pair.
"Balanced"? Seems unlikely. One plate resistor is 15K+15K=30K, the other is 68K||82K=37K, a 24% unbalance.
This type splitter does need a little more gain to the far side of the pair, but often 5%-10%, not 24%. I'd suspect 24% gives a significant unbalance.
Since it does work as pure Pentode with no negative feedback, ouptu impedance is high, speaker damping is low. That will make it sound different on every speaker (and on most speakers, "not as the designer intended"). |
|
|
| beamnet |
The designer is a very respected designer in the Netherlands, his amplifiers are being built in vast numbers by diy enthousiasts. The critics about his amps are usually VERY good.
He himself claims that the phase splitter is very symmetrical, and i know he has a very good set of measuring devices...
Is there a better way to connect the output tubes? Triode? Or even..UL? I have taps, but there is no data available... |
|
|
| SY |
| quote: | Originally posted by PRR
> "this splitter thing" is a well balanced long tailed pair.
"Balanced"? Seems unlikely. One plate resistor is 15K+15K=30K, the other is 68K||82K=37K, a 24% unbalance.
This type splitter does need a little more gain to the far side of the pair, but often 5%-10%, not 24%. I'd suspect 24% gives a significant unbalance.
|
Don't forget the 47K grid resistors in parallel- that swamps the variation quite a bit, bringing it into line with what you expected.
I'll be polite and not suggest that the designer needed to use something closer to a CCS on the tail and matched plate (and o/p grid) resistors... |
|
|
| Johan Potgieter |
Sorry all, I was absent for a while compliments of server troubles.
Firstly, I will try to post the promised KT88 performance when moving G2 from B+ to anode. I have difficulty in scanning the original graph by GEC in 1957, am also careful of copyright infringement, so I have rather tabulated values for 10% steps of G2 "movement". A graph drawn from this would be illustrative.
KT88 TAPPED SCREEN OPERATION
[P.p. operation: RL(a-a) = 6Kohm]
%G2 tap - Pout(W) - Zout(ohm) - D(%)
Pentode ---47-----------20K------3.0
10---------48.3---------12K------2.3
20---------49.7----------8K------1.8
30----------50----------5.5K-----1.8
40----------50----------4.2K-----1.8
50----------47----------4.0K-----1.8
60----------39----------3.8K-----1.8
70----------35----------3.5K-----1.8
80----------30.5--------3.3K-----1.8
90----------26.8--------3.0K-----1.6
Triode-------25----------2.5K-----1.2
From this it should be clear that at 43 - 50% taps the best of both worlds are available. Also note that the distortion (1.8%) does not stay constant; these are the figures at maximum output for each tap, so that at say 25W, the equivalent distortion decreases continually as the max. output rises. The distortion is t.h.d. and the composition is of course important (not given in this example), but I recall from other analyses that for 43% taps it is already largely 2nd harmonic. I am also told that other power tubes behave more or less similar, so that one can take this as a general indication.
From this I cannot bring myself to using either pentode or triode instead. The Zout rises too sharply for pentodes, while the Pout drops too much for triodes. But I would repeat a previous request that I would like to find other data. (Also, as a further illustration, there are curves shown on other sites - pardon my laziness to quote them here for easy access; rather the middle of the night now - which will show that UL characteristics resemble those for triodes rather more than for pentodes.)
About the Schmitt phase inverter (long-tailed pair): Yes, the anode load resistors do differ quite a lot for balance in low mu triodes. There is a practical formula for calculating these if one wants to be academic, which I will not dare try to quote here. It is far easier to sort matters out by experiment. Classically an ECC82 (6SN7) with a mu of about 18 will need the following values for balance: Cathode (common) resistor = 15K; anode resistors = 56K and 68K (21% difference) with next stage grid resistors of 470K. I used this phase inverter preceeded by a directly connected pentode input stage many times in my younger days (sigh . . . ) I usually included a small pot between the RL's for fine balance adjustment. For rather constant working point definition despite valve spreads one could feed the 1st stage G2 from the common cathode (d.c. feedback), but this requires a rather large G2 decoupling capacitor. |
|
|
| SY |
| quote: | | It is far easier to sort matters out by experiment. |
It's even easier to use a CCS for the tail, then matched plate and load resistors. |
|
|
| Johan Potgieter |
PRR - your post #39 (the @#$%*&*& cut-and-past will not work now):
No, sure man, you were not in my face.
This business of what is audible is also - well. . . . a business! It might go a little off-thread, and nerves are easily trodden upon in this field, that is why I stay out. As for how I design, mercifully I never had to vow the public with foofies to make things buyable. I have had enough success keeping the basics right, and staying with "a wire with amplification". But more about that later. Thanks for your contribution. |
|
|
| rdf |
| quote: | Originally posted by Johan Potgieter
KT88 TAPPED SCREEN OPERATION
[P.p. operation: RL(a-a) = 6Kohm]
%G2 tap - Pout(W) - Zout(ohm) - D(%)
Pentode ---47-----------20K------3.0
10---------48.3---------12K------2.3
20---------49.7----------8K------1.8....... |
Hi Johan. I'm missing something here. By tapped screen do you mean ultralinear? I'm trying to reconcile it with the kohms output Z. Where am I going sideways? |
|
|
| Johan Potgieter |
Yes, sorry man - as said I had a problem here and there to get all nicely even.
Under "%G2 tap" the screen goes from 0, i.e. at B+, i.e. pentode which term I used instead of 0%, in 10% steps to 100%, which is at the anode side, where I just said triode instead of 100%.
Thus, e.g. at a G2 tap at 20% from the B+ side, the output under these conditions of current and supply voltage is 49.7W, the Zout (which is actually the internal resistance referred to the primary) is 8K.ohm, at a t.h.d. figure of 1.8% at the quoted 49.7W. At 60% tap (from B+) the maximum output (same supply voltage) will be 39W, the internal resistance 3.8K and the distortion at 39W is 1.8%.
The original graph has as horizontal axis (screen tapping points) 0 - 100%, which is pentode left and triode right, and then vertical axes for the various scales: Distortion up to 5%, output power from 10 - 60 W (suppressed zero) and Zout 0 - 20%, in a convenient way so that curves do not cross each other. They furthermore give 3 output curves for loads of 5K, 6k and 7K. I chose the normal 6K curve; the others follow the same contours. Maximum output into 5K is 57W at 40% tap, and for 7K = 43W.
Hope this is now clear. |
|
|
| rdf |
| quote: | Originally posted by Johan Potgieter
Hope this is now clear. |
99.5% clear. Zout refers to the tube's internal output impedance looking into the primary and GEC used a specially wound output transformer tapped for all these screen drives. The only thing I wish they'ld done is break the THD figure into individual harmonics. There's that big stretch from 20% to 80% tap where the other parameters change dramatically and THD is locked at 1.8%.
Thx for the clarification. |
|
|
| kathodyne |
...i went up from a kathodyne splitter using ecc99 to a long-tail with ccs (was done by a lm317) using 5965
now i have a ccs built with 2xbc547b biased by 2 led's(red)
so the tubes get 5ma each at 150v (similar anode resistors 39k5)
it was so beautiful to see the differences on my scope(kathodyne vs. ccs); even at hard clipping the waveformes stay exactly the same now
it is said bandwidth would now be much better because the lm317 has poor hf specs.
Can one really always draw a straight horizontal line in tube graphs (at the desired ma current) when using a ccs??
if so..
what difference would another type of ccs have (bandwidth?) they all would be almost infinite resistance.
then...why use another type of splitter?? they all have some sort of disadvantage....impedance,voltage swing
is this because of the transistor?? (effectively it should not be in the signal...is it?)
If anyone knows a nice little schematic for tube-ccs (pref. with psu) i'd really like to try!
greets |
|
|
| Johan Potgieter |
Re Rp/RL, previous posts (#35);
This is still on my conscience; I got it from decades back at varsity. It was from a very basic experiment - cannot trace it now. But PRR's advice is sound, so could we just push the Hide button on that one; apology, not intentionally trying to mislead.
CCS in Schmitt phase inverter:
The CCS is in the signal path, and the balance still not strictly perfect impedance wise. The input stage is common cathode with a degenerative cathode impedance in series, the inverter is common grid, cathode driven - higher plate impedance (these are the common objections). But we must get practical; we are dealing with the very limited field of audio. All of this is pretty much academic, with any ill effects way up almost in the RF field. (The same goes for the humble equal Ra-Rc phase splitter.) Tube differences (unfortunately even 2 triodes in the same envelope are not always equal!) are subdued by the degenerative nature of the circuit. The only real disadvantage might be the limited anode swing in cases where maximum is required. A CCS is of course capacity sensitive, but then again it feeds a low impedance cathode drive which swamps that. Oh and of course: one must must keep the cathode-heater voltage in mind (folks would know that it is better to have the heater at a few volts positive wrt an active cathode in case of internal leakage). As said I have never found this phase inverter a problem, even with a cathode resistor instead of the CCS. But in these days of transistors, why not.
I personally would not go for the complexity of a valve CCS, although yes, it SEEMS nicer (a bit of a travesty to put sand devices in a tube circuit. But again let us get with it; this is 2005, and I do not think many still use tube rectifiers.)
Part of being practical must also be consideration of the WHOLE picture. Most of the distortion comes from the power stage; imbalance in an output transformer (and power tubes) probably contributing a whole lot more than a few degrees of phase difference generated by a phase inverter.
. . . . on to the UL data of the KT88. Yes, a break-up of distortion figures would be more informative. But looking at what is available, one would not be far out by taking a (say) 50% tap operation as performing pretty close to triode mode also regarding distortion products. These things do not suddenly jump up and down over the spectrum. At this point the load impedance would have more effect on the final result. Yet again, how I would have liked to have one of those PII audio analysers, but with our exchange rate . . . . I am limited to SPICE, which with the present tube data is not too useful. |
|
|
| SY |
| quote: | | The CCS is in the signal path, and the balance still not strictly perfect impedance wise. |
If the load on each side is the same, I don't see how its source impedance could be unbalanced, at least to the extent that the CS is actually a CS (and it's not hard to make it pretty close to ideal). |
|
|
| Johan Potgieter |
OK SY:
In trying to point out that a lot of criticism from purists in audio is not practically relevant, I was perhaps over-academic in saying "though balance was not perfect . . " - as said then, I referred to the above brigade.
Let as have clarity by an example: For a typical ECC88 circuit with 100K-ohm load resistors, the balance amplitude and output impedance-wise would be within 2% for a common cathode resistance of over 170K-ohm; indeed a very achievable value for a current source, especially semi-conductor. |
|
|
| SY |
| But again, with even a semi-decent current source (say, cascoded bipolars), the balance of output and the balance of source impedance will be completely determined by matching of the load resistors, which is easy to do to better than 0.1%. |
|
|
|
