[Roger A. Modjeski]

As requested here is Primer on OTL amps. I hope to do this without drawing schematics as I don’t think they help the reader. For those who read schematics there are plenty out there. I feel inclined to talk a little about conventional amps for comparison.

First lets consider what the output tube has to do that is different from amps with output transformers. Most tubes want to work at 400-600 volts on the plate, an idle current of 50 mA and see a transformer that is 20-30 to 1 in turns ratio. Whether the amp is single ended (SE) or push pull (PP) the same conditions apply. All SE amps are class A. PP amps can be A, AB1 or AB2 which is just a matter of how much idle current they run. Note I did not say bias. Bias is the negative voltage on the grid as I have said before.

A typical 25 to 1 output transformer steps down the voltage by 25 and steps up the current by 25. People tend to think mostly about the voltage but the current is just as if not more important.  People talk a a lot about matching the impedance of the speaker to the impedance of the tubes. This is foolishness. The tubes don't have a desired impedance, that is determined by the voltage and currents in the circuit. The typical load for a EL84 is 8,000 ohms however the transformer in the RM-10 is 13,000 ohms for the same tubes, but the voltages and currents are quite different. The difference allows me to get 35 watts out of a pair safely.

On the matter of impedance, the output tube I am using has an impedance of thousands of ohms but it is more capable of driving a 8 ohm load than the 6AS7 which is just a few hundred ohms. So I hope you can see that impedance matching doesn't matter. If you start reading a paper about matching impedance note that it only affects damping and is usually not low enough to provide adequate damping without feedback. Feedback is necessary in OTL amplifiers for them to have decent performance. Feedback is not evil if it is properly done. The advantage of OTL amps is that there is no transformer in the feedback loop and, if the loop is short as mine is, feedback is easily applied and not problematic. People who foolishly add 2 dB of feedback to prove feedback harms the sound as I have read twice now in Stereophile reviews, have apparently added it poorly. Why even bother with 2 dB. It takes 6 dB to cut distortion is half and double damping. Two dB does nothing.

To make a good OTL you need a tube that can conduct appreciable current close to one amp and not be damaged in the process. The 6AS7 will pass one amp but little sparks come off the cathode like a 4th of July sparkler. Those cathode flakes rapidly reduce the emission and thus the current. I recently tested a set of 16 tubes for an Atmasphere OTL. They looked physically great. Good getters, no sign of overheating, yet they had less than 40% emission. That means 40% current. I assume that heavy play had sparked off 60% of the cathode material. Horizontal output tubes (mistakenly called "video tubes") were designed to conduct high peak currents and do not shed cathode material even at 1.5 amps. So I use those tubes.

Now one must arrange the tubes in some kind of circuit that connects them directly to the load. Since transformers are out standard Push Pull is out also as it requires a center tapped transformer. What is left is to either stack the tubes or put them in a bridge. While one can make a single ended OTL, that is foolish as the idle current has to equal the output current. The dissipation in the tube would be too large.

The SEPP (single ended push-pull) is another obvious choice but the self bias resistor is a problem. My direct drive ESL amp is SEPP but it runs at 2500-5000 volts. SEPP amps work great up there but are worthless at voltages needed for 8 ohm speakers. An SEPP is a stacked tube output circuit where the top tube is self driven so that one only needs drive the bottom tube. No phase inverter is needed. Its a great circuit when the parameters are right but they are not for a 8 ohm speaker.

Now we have it down to two circuits. The totem pole (Futterman) has a top tube that pulls the speaker up positive and a bottom tube that pulls it negative. This is exactly what happens in a standard transistor amplifier. However most SS amps have P and N complementary transistors, we have only N type tubes. Therefore we have to figure a way to drive the two tubes equally even though the top and bottom tubes have very different drive and DC requirements.

The main accusation leveled at totem pole outputs is that the top tube never gets the proper drive which is equal to the bottom drive plus the output voltage which is large and varies with the load. If the Futterman connection is looked at casually that appears to be the case and many assume that the Futterman uses large amounts of feedback to correct this situation. However, a properly constructed Futterman circuit has equal drive with no feedback at all. What I notice in Rozenblit, Atmasphere and other writers is they first have to damn the Futterman to present their "correct drive" which sometimes is not correct at all. I see these mistakes all over the internet by somewhat respected sources.

Perhaps they do understand the circuit and hope you won't be able to argue their points. (many Japanese authors have joined them in anti-Futterman articles). Who knows enough to contest their technical argument especially when it's in a respected magazine like Glass Audio. Last I looked Rozenblit had some complex way of driving the top tube differently than Futterman. So his amp goes into the totem pole camp although now I see Single Ended amps are prevalent on Trancendent's website. One uses a 300B which makes no sense at all as it does not like delivering current. We would not want to see our EML 300Bs in that amp as they wouldn't be there for long. Some of these less than good ideas get past some listeners who listen at very low levels. Those who push them usually find tube life to be that of a Mayfly. Here is a typical Rozenblit analysis of the Futterman, it is entirely untrue. http://www.stereophile.com/content/transcendent-t8-otl-monoblock-power-amplifier-futterman-redux. Unfortunately Wikipedia does not have a good article on the Futterman.

The other configuration of OTL is taken from the Electrovoice Wiggins Circlotron  http://en.wikipedia.org/wiki/Circlotron. This brief article is worth reading and has a schematic and some interesting numbers. Karsten's story, as he once told me, is that he knew of the Circlotron and wanted to eliminate the center tapped output transformer. "One day while napping on the couch it came to me how to do it" Not much of a story because all he did was leave it out. The circuit is virtually the same. I believe the original EV driver had some bootstrapping which I am not sure if Ralph uses now or not. In any event you can see it takes 2 or 3 dual triodes to drive the output tubes. My circuit does it with one 6GH8 pentode/triode in a very short path, DC coupled internally. Futterman had internal cap coupling and a big one on the output which limited the low end. The last thing we want is a big electrolytic between our amplifier and speaker.

So that's about it. Questions?

[steve f]

Roger, could you elaborate a bit concerning the "mistake" Futterman made in driving the top tube in his totem pole arrangement? The Transcendent circuit is complex. The high distortion numbers reported in the Atmasphere type of amps aren't very appealing. What is everybody missing concerning Futterman's work?

Many OTL amps suffer from poor tube selection. Some sound good, but tube failure rates are high. Is this par for the course in QC, or are these tubes being used outside their parameters?

Thank you,
Steve

[Roger A. Modjeski]

Roger, could you elaborate a bit concerning the "mistake" Futterman made in driving the top tube in his totem pole arrangement? The Transcendent circuit is complex. The high distortion numbers reported in the Atmasphere type of amps aren't very appealing. What is everybody missing concerning Futterman's work?

Many OTL amps suffer from poor tube selection. Some sound good, but tube failure rates are high. Is this par for the course in QC, or are these tubes being used outside their parameters?

Thank you,
Steve

Futterman did not  make a "mistake". The mistake is being made by those incorrectly analyzing his circuit. The OTL-3 is quite complex and a casual look does not reveal its true genius. I had to look for a long time to see what they had missed.

On your second point, I would say many OTL amps suffer from poor tube choice. To me selection means finding the proper tubes out of a batch and matching them.

The 6C33 and 6AS7 tubes used by most OTL makers today were not meant for this application, they are a bad choice. They are plentiful and very inexpensive. They are very low mu tubes (around 2) that were intended as series pass tubes in vacuum tube power supplies. They are not designed for the high peak currents that audio amplifiers must provide on music peaks and transients. I saw this easily when curve tracing these tubes. The curve tracer pulses the tube at 120 HZ up to its saturation current. Even with a 300 ohm load several sparks fly off on each pulse. This is likely because the grid is so close to the cathode that the emission is not well distributed. These cathode sparks are bits of cathode material. As this cathode material sparks off the cathode looses emission and can produce even less current. Perhaps owners have noticed that after playing their amps loudly they don't sound as good the next day. Party on. but don't turn it up too much.

Now lets talk about matching. Atmasphere claims you do not have to match the tubes. I disagree. I had the opportunity to work on a pair of MA-1s which use twelve 6AS7 tubes per side. The customer was having random tube failures. I put in 12 tested but un-matched 6AS7 tubes and measured the cathode current of each. I found some were conducting half as much as others which means others were conducting too much. The tube carrying the heavy current would of course wear out and fail first. By matching the tubes I solved this owner's problem which he had been dealing with for years using unmatched tubes.

A similar thing will happen if there is loss of emission as in the second amp I re-tubed. A few tubes were still at 90% emission but some were as low as 20%. In that case the 90% ones were carrying almost all the current which was intended to be spread equally among the tubes. Note that unless the user can verify the condition of each tube he will simply be re-biasing the amp over time and the strong tubes will have to carry more current than they did initially because the weak ones are doing very little of the work.

[Maritan]

Roger - What are your thoughts on David Berning's ZOTL technology? Seems interesting, but I haven't had enough time to delve into it much.

[Roger A. Modjeski]

Roger - What are your thoughts on David Berning's ZOTL technology? Seems interesting, but I haven't had enough time to delve into it much.

David is a smart guy and often does unusual things. From what I can gather the amp is running at RF frequencies and is not a conventional OTL as we are discussing here. Good luck on understanding this

From his website I copy the following portion: Here is the link for the page http://davidberning.com/products/zh_211_845

The Berning ZOTL amplifiers take advantage of new technology in order to eliminate the audio-output transformer. These amplifiers use radio frequency to change the voltage-current transfer characteristics of the output tubes from their normal impedance-plane to one suitable for driving a dynamic loudspeaker. The radio frequency re-mapping is implemented using special high-frequency power conversion techniques. The high-voltage, low-current tube impedance-plane is re-mapped to the high-current speaker impedance-plane through a special transformer at a constant RF carrier frequency of typically 250 kHz. Because the audio signal is riding on a carrier, it is not subject to parasitic elements of the transformer that would otherwise distort the audio signal. There is no low-frequency limit for this impedance conversion, and these amplifiers are dc-coupled, with the exception of a small, high-quality, dc-blocking capacitor at the input of the amplifier to prevent the amplification and application to the speaker of any dc that might be present at the input.

The impedance matching in the ZH amplifier is precisely determined by the effective turns ratio of the RF conversion transformers. Both the push and pull portions of the amplifier are handled in a symmetric fashion, as they are in a conventional transformer-coupled amplifier, and negative feedback is not required to force the positive-output current capability to be similar to the negative-output current capability.

[atmasphere]

It appears that some things need clarification.

Roger has the story about how I arrived at the Atma-Sphere circuit incorrect. It is a bit apocryphal. It is true that the inspiration came to me while dozing off but there was a lot more to it than that, most of which occurred in my waking hours  . The thing about the center tap just never happened.

Regarding 6AS7 reliability and the need for matching: Roger did indeed match a set of tubes for us, if he did it for someone else that event was never recorded. The set that he matched for us was not done for reliability purposes. We placed them in an amplifier that got a review, and the fact that they were there was noted by the reviewer. We have not seen any correlation with matching and reliability whatsoever.

We warranty our 6AS7Gs for one year. We have been in business for over 38 years; those two facts should tell you all you need to know about reliability, as it has been reliability issues that have put most OTL manufacturers out of business.

Regarding the Futterman circuit: while Julius apparently did a good job with the amp, succeeding designers wishing to cash in on his circuit did not, and in the process created what we have come to call the 'Futterman Legacy'. In the old days when people would find out that we made an OTL, they assumed that it blew up... we didn't have go bad-mouthing Futterman as Roger suggests- the amps themselves did that pretty well on their own. In particular, Harvey Rosenburg, who assumed the Futterman mantle, had enough reliability problems that he pretty well convinced the world that OTLs blew up- in a way that was much more convincing than any analog/digital or tubes/transistors debates did for their proponents.

(The fact of the matter is we created the first reliable OTL that was made in any appreciable numbers. It has no bias drift issues (never did) and also has instantaneous overload recovery (and always has). Those two issues have been and continue to be killer problems for a lot of OTL circuits.)

Actually the real problem was not so much the Futterman circuit but execution! The Fourier amps used the Futterman circuit, and blew up quite a lot, not so much because of the circuit but because filter caps of questionable origin were being operated 40 volts past their maximum rating! Harvey Rosenburg liked to use surplus photoflash capacitors, they sounded nice but had poor ripple current specs and thus did not hold up well. When his source on those dried up it put at least one of his models out of production.

Reliability has driven most of our competition out of business. Its really all about economics- if you have to repair every unit you make several times you aren't going to last long.

Regarding feedback: Of course feedback is something a lot of OTL designers want to use. We have either used none or very small amounts as we have a particular goal: we want the result to sound like real music. Its a simple fact that if larger amounts of feedback are used, the result will be an artificial brightness brought on by trace amounts of odd ordered harmonic distortion (no amplifier has a signal path so short that it can bypass this problem). This is why two amps can have the same bandwidth (we get about 300KHz out of our Mk 3.2 series) on the bench but one might sound bright and the other does not. The idea of using feedback is to try to create a voltage source out of the amplifier and this can be done with an OTL. You want the constant voltage response so that the amplifier will have predictable flat frequency response on your loudspeaker of choice.

The problem is that the use of feedback causes the amplifier to violate a fundamental rule of human hearing: how we detect sound pressure. When those trace amounts of odd ordered harmonics distortions (which are often so low that they are difficult to measure) appear, the ear/brain system interprets them as brightness, and also uses them to assign a higher volume level to the sound. When they are absent, in essence you can play the stereo a lot louder without fatigue.

In high end audio this is the difference between sounding like a nice stereo and sounding real.

If your circuit employs little or no feedback, accurate frequency response may not be in the cards. I say 'may' because not all speakers require a perfect constant voltage response out of the amplifier. In fact a good number of speakers today work better if the amp provides a constant power response instead. This is true of nearly any speaker than can be driven by an SET. You get a constant power response out of the amp if it has little or no feedback. We chose this course because we are not interested in speakers that have no chance of sounding real.

The ear/brain system interprets distortion as tonality. It has tipping points where the distortion can be favoured over actual frequency response errors. All speakers have frequency response errors anyway- the fact of the matter is that a pure constant voltage response is really not needed. If it were, transistors would rule the day and there would be no tube amps at all. But there are tubes, the economics have shown that there must be a good reason.

Admittedly this has limited our market for a long time. But it has also created a very loyal customer base, as once the amp is set up correctly, its going to be a tough job to beat it with anything else on that speaker. The limit in the market may well be the reason we make the only production OTLs without feedback.

[Roger A. Modjeski]

Ralph,

Thanks for correcting the record on several points and being a gentleman about it. You have made a reliable amp from what I hear. I had forgotten I matched up a set of tubes for you, that must have been a long time ago. The matched set I referred to was for a customer in San Francisco who sent me the amps and came to pick them up in person. You helped me with the cathode resistor upgrade on this older amp. With the cathode resistors in place I could measure the current in each tube and the differences in unmatched tubes did concern me. From the numbers I obtained it was easy to see there was a broad range of currents. I know you like to say that matched tubes are not required but I feel the tubes will run longer if they are sharing the work.

I agree that NYAL (Harey's company) and Fourier contributed to the Futterman legacy. Keep in mind those amps were completely on a PC board which didn't help. I have worked on a Fourier and had to re-design it. Yes the caps were being over-voltaged. The NYAL amps have a reputation for oscillating which is entierly a question of layout and feedback compensation. Even though they were building Futterman's circuit they were not building Futterman's amp.. The amps Futterman made himself appear to be quite reliable.

On the subject of constant power I cannot agree. As the impedance of many speakers rises above 40 ohms at the low end resonance, won't the voltage go up several dB and the SPL with it? When we see the impedance dips due to the port and crossover, won't the SPL fall several dB? I find that frequency response differences are the first thing we hear when we change amps in a system we know well. Although the new amp may have better detail or more air, what are we to do about the altered frequency response and the flabby bass?

On the topic of feedback the shorter the path the better and this amp has a very short path with only one coupling time constant in the loop.  In my OTL the input tube literally stands on the output rail. The -3 dB down point is half a megahertz. Therefore there is no delay at audio frequencies.

Perhaps I am shooting myself in the foot referencing Futterman at all. Perhaps we should just say this is Roger's amp and it has a totem-pole output stage which differentiates it from your amp which has is a bridge.

Is the quote from you accurate: "The Futterman circuit should have never seen the light of day"?

[steve f]

Roger,

I agree that choice is the better term. My bad.  Of course my use of Futterman's "mistake" in quotes means not to be taken literally. His circuit is above my current level of understanding.  I am surprised that knowledgable people with EE credentials dismiss his ideas. They really should not group subsequent variations with the original.

When are we going to get a glimpse of Roger's OTL?

Ralph,

Would you be so kind as to list a few speakers that match well with your product?

Steve

[atmasphere]

I have a paper on our site regarding this topic:

http://www.atma-sphere.com/Resources/Paradigms_in_Amplifier_Design.php

Its all about intention. If the designer of the speaker expected a higher output impedance, the impedance peaks in the bass may not be a problem at all, instead may simply represent another half octave of bass extension, without boominess.

Voltage Paradigm rules were created 50 years ago before human hearing/perceptual rules were really understood. Its telling that they have not been updated in that time- human hearing is a bit- inconvenient.

We built an OTL with 10MHz bandwidth and it still was not fast enough that the negative effects of feedback could not be heard. What is important here is understanding that the ear/brain system uses higher ordered harmonics and is really sensitive to their presence.   The ear/brain system is also tuned to birdsong frequencies from millions of years of evolution (look at Fletcher Munson curves) and many of those slight harmonics occur in the same range, and not by coincidence.

To get away from this problem you have to dump feedback.

Its no big shakes to have problems with low impedances. There is no argument for low impedance speakers if sound quality is your goal. All amplifiers, tube or solid state, make more of the wrong distortions driving lower impedances. BTW, increasing feedback does not affect output impedance. To do so violates Kirchoff's law.

Duke Lejuerne of Audiokinesis (audiokinesis.com) can and has explained how the impedance bump of a woofer in a box does not have to be a problem. If you do the math, you find that many speakers will work with a medium high output impedance with minimal coloration. What is often more important due to tipping points in the brain's hearing/perceptual system is keeping the audible distortions (like intermodulations and higher ordered harmonics) to a minimum, as it will convert those to tonality. That 2nd order harmonic that gives so many tube amp warmth is a well-known example! IOW, its all about intention.

[atmasphere]

Ralph,

Would you be so kind as to list a few speakers that match well with your product?

Steve

OTLs have an economy of scale. Our S-30 works fine with the Rogers LS3-5A, Quad 57, most horns and full-range single driver speakers, Merlins, and most speakers where 30 watts is enough power and the impedance in the bass is 8 ohms or more. We play them with High Emotion Audio Bella Twins here at the shop. Zus are also a good combination. The S-30 was really intended as a smaller amp to take on SETs, usually on their own turf, meaning some sort of more efficient speaker. But it has more power than most SETs, so it can be used on a wider range of speakers.

The M-60 can use a larger pantheon and is often seen with Quad 63s and later (and there are a lot in use with ZEROs as well) and works with many Wilsons, but I think the Merlins are used with M-60 more than most any other speaker.  MA-1s can drive Sound Labs, , Wilsons, Karma, JM Labs, Spendors, to name a few. MA-2s do that better and don't run out of power even on MBL 101bs (although a Zobel Network is recommended to prevent midrange brightness). Apparently if you want bass out of Avalon Eidelons the MA-1 is one of the few games in town.

The MA-3 is the biggest OTL made and can drive most speakers around. The bigger amps are quite nimble and show no signs of sluggish behavior that is common in larger tube amps. That is something OTLs do really well- they are usually very fast even in higher power versions.