Go to Atma-Sphere's home page and read Ralph Karsten's white papers on the subject. You can also contact him directly by mail. He is a nice and very helpful person.
14 responses Add your response
You could also wait until I find this thread :)
I'm not sure what you mean by 'OTL amp get its power'... of course the power comes from the wall :) seriously though an OTL amplifier is able to make its power in a way that is
not unlike a transistor amplifier. A lot depends on the type of tube used, some tubes are vastly more suited than others, mostly due to a low plate resistance and high transconductance.
The primary advantage of an OTL is that without an output transformer, distortion is reduced and bandwidth is increased. A less obvious but still important advantage is that an OTL can be a simpler circuit, as with any output transformer the output voltage of the tubes has to be stepped down to loudspeaker voltage, whereas in an OTL this does not happen, so you don't need as many gain stages in the amp. In our case, that means there is only one stage of gain, making for a fairly simple signal path. The less stages of gain, the more bandwidth and lower distortion.
Anytime you reduce distortion, the result is a more detailed sound that is simultaneously smoother. Increasing bandwidth can have the effect of increased impact on the bottom with greater low frequency extension (although right here I will interject that so-called 'tight bass' does not exist in the real world and is a phenomena of excessive negative feedback in an amplifier design), and an obvious increase in speed on top.
The heat is a function of the the class of operation, just like with any other amplifier. A class A amplifier will run hotter, regardless of the technology.
Clipping is a function of the power of the amp. Some OTLs can be unstable at clipping, but that can be true of many conventional amplifiers too. We have built OTL guitar amps that are intended to be overdriven and they work quite well. I believe that any proper amplifier design, regardless of technology will have instantaneous overload recovery and will be unconditionally stable- that is to say it will be stable regardless of the input signal or output load.
'Reserve power' is a term that refers to the class of operation- by definition a class A amplifier will have 0 db of reserve power. IOW, the better the amplifier (regardless of whether it is an OTL or not) the lower the reserve power figure will be (FWIW this term is counter intuitive on purpose to make less expensive AB amplifiers look better).
OTLs can drive JM Labs speakers quite well. This has more to do with the power of the amplifier rather than its technology. JM Labs speakers, IMO, have traditionally been tube-friendly, but they do need some power.
I agree with virtually every point Ralph raises, but as everything in life, it goes both ways . . .
Advantages of OTLs:
1. No output transformer. Really, this is THE advantage and rasion d'etre of OTL amps, hence their name (refers to something they don't have). Whether the performance attributes of audio transformers offset their disadvantages is of course a multi-decade debate.
2. Gain efficiency - actually, really something I hadn't thought of until Ralph mentioned it.
3. Higher potential slew performance.
Disadvantages of OTLs:
1. The transconductance characteristics of vacuum tubes operated in an OTL push-pull fashion is both inherently non-conjugate and non-complimentary - essentially similar to a the "all-NPN" solid-state amplifier designs of the early-1970s. Class-A biasing helps tremendously, but this will always be a fundamental source of large-signal even-order non-linearity, even at higher harmonics. A tranformer-coupled push-pull topology is still non-cojugate, but is inherently complimentary, and provides reliable cancellation of even-order distortion.
2. The plate resistance of virtually all vacuum tubes is WAY too high for effecient power transfer to a typical loudspeaker load. Paralleling a bunch of output tubes is the usual solution, and power-efficiency of OTLs is still very poor, even worse with all of those filaments to run. Now when direct-coupling to electrostatics, it's a whole different story . . .
3. OTLs may be gain-efficient, but they're definately NOT voltage-efficient, and require split high-voltage power supplys (or capacitive coupling, but then what's the point?). The primary inductance of a transformer, in contrast, makes for a VERY efficient use of power-supply voltage, as the maximum AC voltage peaks can be much higher than the B+.
4. Vacuum-tubes have comparatively poor DC-offset performance, and while solveable, this can present significant engineering hurdles. Conventional transformer-coupled topologies (should be) inherently DC stable.
5. The poor power-transfer-efficiency eats up virtually any possible advantage in slew performance, and stability issues for application of global NFB are identical for both types.
In the end, it's obvious that transformers are inherently imperfect . . . but also obvious that a perfect transformer could solve SO many engineering problems. Virtually every aspect of building an acceptable OTL amplifer involves huge sacrifices in efficiency . . . and (but?) efficiency isn't that great with ANY tube amplifier to begin with . . .
Any way you slice it, there are some significant obstacles to making the perfect amplifier.
Are you saying the step up transformers in Quads can be bypassed?Actually, I meant that remark in the sense that there's nothing inherently "wrong" with the fact that tubes have relatively high impedance characteristics . . . it's just very different than the low impedances of dynamic loudspeakers.
For a high-impedance (i.e. electrostatic) loudspeaker, then there are indeed methods for directly coupling a tube amp to its electrodes, and a very efficient transfer of power is possible . . . Acoustat did this in the late-1970s, with very well-received, albiet unreliable results . . . I serviced a few of these many years ago, and can personally attest to both of these characteristics.
A direct-drive electrostat these days is still a pretty intimidating engineering project - the best common tubes suitable for the task (i.e. TV sweep tubes) are long out of production, and custom transformer(s) will probably still be required, for the high-voltage supply. And if you're talking about the ESL-63s, there's the bit about the multi-tapped delay line that would probably be a complicating factor.
If I wanted to build an active electrostatic speaker with a tube amp, I'd probably look more to using a conventional push-pull output stage with conventional, readily-available audio tubes, with a special low-ratio step-UP output transformer to match the electrostatic panel. But it'd still be a LOT of work to get it right.
While Athmaspere OTL amplifiers are superb, and no question about it! I still wish to have fairness to other OTL designers who do not participate in on-line forums. One of them is Jud Barber of Joule-Electra. Not many audio designers have so many truly prestifeound awards like Golden Year by Harry Pearson as Jud has.
His web site is not maintain very well so you have to do a google search. Below is review of of David Robinson, The Editor-In-Chief of Positive Feedback
In no way I am saying that one designer's equipment is better then another. Its all up to the listener to judge.
Rafael, no I do not agree with all his comments, although at at an earlier time some were true.
For example, the comment
1. The transconductance characteristics of vacuum tubes operated in an OTL push-pull fashion is both inherently non-conjugate and non-complimentary - essentially similar to a the "all-NPN" solid-state amplifier designs of the early-1970s.
applies to Futterman amplifiers only. The Circlotron output circuit (we were the first to use this in a practical real-world OTL) eliminates this problem in both tube and transistor circuits, allowing one to use non-complementary pairs (and for the record, complementary transistor pairs, such as NPN and PNP are never exact matches, so the argument is really a red herring).
I also have to clarify something about this statement:
2. The plate resistance of virtually all vacuum tubes is WAY too high for efficient power transfer to a typical loudspeaker load. Paralleling a bunch of output tubes is the usual solution, and power-efficiency of OTLs is still very poor, even worse with all of those filaments to run. Now when direct-coupling to electrostatics, it's a whole different story . . .
There *are* tubes that have low plate resistances. The 6AS7G is an example, as is the 6C33. So in an OTL, you will not find anyone using 6550s or EL34s! It is true that you still have to parallel tubes, but of any higher power amp that is a fact of life. Kirkus is correct about the filament issue, although the filaments often get blamed for excess heat, which they are *not* responsible for. That comes from the class of operation.
Comment #3 applies to Futtermans OTLs only. Circlotrons don't use a split-voltage supply, nor is there any need for an output coupling capacitor. Many people think that if its an OTL, it has to be set up like the old Futtermans, which do have these 'design features' but that is not true.
Comment #4... not IME; many customers of ours have commented on the fact that they can set the DC Offset of the amplifier and it will be exactly right 6 months later. The trick (and you would think this is obvious) is to be able to control the power tubes.
Comment #5 was never true- even the old Futterman amplifiers from the early 60s had slew rates far in excess of their transformer-coupled counterparts. We've measured slew rates of 600V/micro-second in the output section of our amps. This translates to extremely wide bandwidth. Our early prototypes exhibited this trait right away- they made very capable RF booster amplifiers at frequencies as high as 50MHz without oscillation. In our production amps we limit the bandwidth in the driver circuit to minimize RF issues, but the output section retains its speed.
Dear Ralph, As usual your comments are extraordinarely appreciated. I truly thankful to you for your (at very least) educational role for me and I am sure many other members of this forum.
To my regret, I never in my life auditioned your amplfifiers and only a few weeks ago became introduced to OTL amps at all (what a wonderful experience - I could not stop listening to music.... and I auditioned amplifier or two in my life). The efffortlessness, the "magic", the...
.(no words) ... You and Jud Barber are magicians !!!!!!
Thank you again,
I actually tried to speak very generically about some of some of the challenges associated with designing a successful OTL amplifier . . . which seemed to be somewhat relevant to the original poster's question. I was NOT referring to disadvantages with Atma-Sphere in particular. NONE of these are necessarily of any particular disadvantage to the end user of a competently-designed amplifier, but they ARE obstacles for the circuit designer.
I personally have a little experience with the Futtermans, and am reasonably familiar with the topology-usually-refered-to-as-"Circlotron" (which isn't necessarily/originally an OTL) . . . and these circuits REALLY aren't all that different from each other -- they're all variations on "push-pull". An excellent analogy would be the QSC solid-state amp vs. the conventional topology - it's disorienting to look at the schematic, but all of the same elements are still there, doing the same things. I share Ralph's view of the "Circlotron" as being the more elegant arrangement, mainly because of the equal-amplitude drive voltages. And I'd speculate that we might agree that the capacitor-coupled "totem-pole" arrangement as being the most problematic.
Engineering is very much an "in-spite-of"/"because-of" kind of discipline . . . and IMO it's the ability to keep this in balance that defines whether or not a design is ultimately successful. Atma-Sphere's longevity as a company is a strong testment to their product being hugely, vastly improved over the NYAL (Futterman) amplifier, and they get good reviews for sound quality in-spite-of/because-of (choose one) the fact that it's OTL, tube, low-feedback, class-a, high-output-impedance, etc. etc. etc.