SET vs OTL

My current SET amp has been 100% reliable since I got it over 5 years ago and is the most quiet tube amp that I have ever owned.

Reliability is really important! When we went into business, it was in the face of OTLs being the least reliable kind of amp made. But we sorted out how to make them hold together, which is what has kept us in business the last 46 years.

How about an OTL SET ?

Its certainly doable and has been done. But if you are the kind of purist that wants only one power tube its not going to happen. The most likely candidate (Transcendent Sound) makes 1.5 watts using four EL509s. IIRC the way this is done involves using the grid of the power tube as the output means.

I know Berning is not strictly OTL but from what I understand the transformer is only there to filter some very high frequency noise riding the signal.

This is incorrect- the output transformer is passing the audio signal just as in any amplifier. But its a bit different as the output is used to load an unregulated 250KHz power supply and so the audio is imposed on that frequency. The 250KHz frequency is then filtered out at the speaker terminals.

2. Superiority is a given or an absolute

This is correct.


IMO/IME the OP question is a bit too broad. Some OTLs sound more like SETs (like ours, due to no feedback and otherwise masked higher ordered distortion content); some sound more like traditional PP tube amps (due to feedback, combined single-ended and PP circuitry, and the use of pentode output tubes).


All these things make and enormous difference in an OTL because there's no transformer to hide what's going on.


On the SET side, you have traditional SETs, parafeed, Loftin-White circuits; Jack Elliano developed a class A3 output system (which he patented; it seems that Western Electric is using his system) and believe it or not some SETs employ feedback. They all sound different, and they also all sound different depending on whether you have a type 45 output (arguably the ’best’ since they have the widest bandwidth although they are not the best in that they have almost no power), 2A3, 300b, 211 and so on, and whether there are paralleled output tubes and of course how well or poorly the output transformer is designed.


Its also important that if you are doing a comparison, that both amps be comfortable on the loudspeaker chosen for such a comparison. Smaller OTLs need higher impedance speakers, but if you want the best out of any SET, higher impedance (16 ohms) works out well in their favor as well! I think @pani did a good comparison between his amp and the Tenor on this latter account. But while the Tenor used many of the concepts we pioneered (fully differential voltage amplifier, Circlotron output stage) it was a *very* different sounding amp from our stuff.


So I agree you can’t make blanket statements, but IMO/IME you can indeed make very specific statements! For example I’d put up our S-30 against any of the more expensive Audio Note amps out of Japan and as long as both amps were comfortable with the speaker load, expect the S-30 to be better in every way that can be described. But there are some speakers that the Audio Note amps can drive that the S-30 cannot (to be specific: 4 ohms) and in those comparisons the S-30 would fall flat on its face. So I imagine that its important to pay attention to what the designer of the amp had in mind when it comes to loudspeakers.


IMO/IME the distortion signature of any amp is far more important than what kind of amp it is. If the amp has, as its primary distortion product, enough of the lower harmonics to mask the presence of the higher ordered harmonics, you’ll get that organic musical sound that so many SET users crave and it won’t matter if the amp is SET, OTL, solid state or class D.

And this is even in systems which are super efficient and made for SETs (single drivers, Audio note speakers etc). SETs sound like "they are here" whereas push-pull sound like "we are there". At any volume and with any speaker I have heard, this is another phenomenon that was consistent

People occasionally ask me how a stereo is supposed to work, and I explain that the best model I've come up with is that the stereo can act like a time/space machine that can graft your room onto a musical event.


Some recordings are very intimate and sound like they are in the room. Others are larger events, like a full orchestra, in which case you feel more like 'you are there' rather than 'they are here'. IOW IME the way that presents really depends more on the recording if the stereo's ducks are in a row.

That is also one reason SETs sound more vibrant & energetic compared to Push-Pulls.  

Actually what is happening is how the SET makes distortion. The energetic quality is due to higher ordered harmonics showing up on the transients when you ask the amp to make any significant power. Those higher orders are interpreted by the ear as 'louder' although if you were to use a sound level pressure meter, you'd find that its an illusion. They are a sign that the speaker isn't efficient enough to take advantage of the better properties of the amp. Any time you ask an SET to make more than about 20-25% of full power these higher ordered harmonics are in the mix.

That speaker seems a bit on the low side of efficiency (93dB) for an amp of only 7-10 watts! Nice speaker though. I'd be putting at least 10x more power on that unless the speaker was in a small room.

This is what I know about the Tenor (we had a few in here for service some years back): It is a differential design but using a circuit similar to what is in the BAT VK-60, which is to say two cascaded gain stages driving the output section. That is how our very early amps worked but back in the early 1980s I realized that there were too many problems with the approach (in particular DC Offsets which would result in unequal drive to whatever stage was succeeding, resulting in distortion, especially the 2nd harmonic).

The 2nd harmonic suggests a quadratic non-linearity is present. The differential design suggests a cubic will be present also. Further that amp employs feedback (although not nearly as much as is needed) while the WAVAC does not. I would expect the Tenor to be fast and transparent but a bit dry, perhaps a bit bright and less engaging compared to the SET. Do I have this right?

@pani Which Tannoys did/do you have?

Ralph, are you able to address my comparison between Graaf GM20 and Wavac EC300b in the same way? I am not yet convinced that the issues raised by @pani aren‘t valid in the same way.

From what I can see of the Graaf front end circuit, while it is balanced it must have a fairly low CMRR; this based on the lack of a constant current source. This causes the circuit to have higher offsets, lower gain and higher distortion. This in turn muddies the water as far as the distortion product goes (mostly due to the offsets generated, which cause more 2nd harmonic to be created)- you can't say its entirely cubic insofar as its non-linearities are concerned. So I can't really comment about the comparison and certainly can't contest your findings! But of course the Graaf is one of a number of OTLs out there.

Regarding the balanced connection thing, I always encourage people to look at whether the gear doing the balanced connection supports the balanced standard or not. The reason is simple: if you don't support the standard you don't accrue the benefits either! FWIW most high end audio manufacturers don't seem to support the standard. One of the more important aspects is that in balanced mode, neither the inverted or non-inverted signal references ground- they reference each other and ground is ignored. Its this bit that makes supporting the balanced standard a bit tricky- most of the time you need an output transformer to do the job. To do that right is expensive.We patented a means to get around that problem but that simply means we're the only one doing it our way, which is direct-coupled.

I have heard the exact same distortion attribute when I compared:

1. Audionote SET to Leben push-pull
2. Ayon Crossfire II SET to Triton II push-pull
3. Trafomatic Experience SET to Traformatic 6550 pushpull
4. Wavac EC-300b SET to Canary audio 300b push pull
5. Border patrol 300b SET to Border patrol 300b push pull
6. Unison research Kt88 SET to CJ Premier KT88 amps
6. The list goes on...
All of them were side by side comparison with my own music.

As long as we are talking about bandwidth, power, detail, THD, we are not going to meet anywhere.

@pani  You may have noticed that I've been talking about *distortion signature*, which isn't the same as THD. You can hear these things very easily- quite literally its the difference you're describing. The reason bandwidth is important is that if the amp is using little or no feedback, bandwidth is needed to preserve phase relationships. Detail is a subjective idea so probably isn't measurable, but we do know that it can be masked by louder sounds as that is one of the human hearing perceptual rules.

Regarding the above list, every. single. one. of them mix single-ended circuits with balanced (PP) outputs. This results in that 5th harmonic prominence I mentioned earlier; IOW  you're making my point.


Its simply erroneous to assume that because all the amps you've heard don't bring home the bacon that *all* PP amps won't. There are detractors of class D amps that make the same mistake- because they heard several class D amps that sounded bad to them, the assumption is that all are bad. In this case I'm making the statement that you need to compare a fully differential circuit against one that is single ended only. You've not done that yet so far as I can see.

"it feels like a one sound" I get exactly what he means. This is what I recognized when I got my SET and listened compared to my otherwise very push pull amplifiers.

Funny - people describe that about our stuff too.


As best I can make out, what seems to be important in any amplifier is the distortion signature, possibly more important than *how much* distortion it actually has, although the lower you can make the distortion the more detail you can reveal. The important thing in the distortion signature seems to be that the lower orders be in sufficient amount that they can mask the higher orders. Many solid state amps violate this idea by having very little of the lower orders, leaving only small amounts of the higher orders. Because the ear is so sensitive to those orders, they are easily heard, causing such amps to sound harsh and bright, especially at higher volumes. This is literally why tubes are still about 60 years after solid state was introduced to hifi.


I recently built a little 5W/channel tube amp using output transformers. The power tubes are EL95s which are a cute little pentode power tube. I wanted something that was low power, compact (as an integrated stereo amp this one can sit on a sheet of paper with full margins visible) and otherwise as high quality as I could come up with (it was for my bedroom system and also a high quality desktop). It uses a differential amplifier for the voltage amplifier/driver circuit (IOW no separate 'phase splitter'). The output section is ultralinear and class A. A high quality CCS circuit was used for the 12AT7 input differential amplifier, so its CMRR numbers are quite high. So literally two power tubes and a 12AT7 per channel. I gave it two inputs and a volume control.


The inputs are single-ended although this amp is fully balanced/differential from input to output. The other input to the differential input voltage amplifier is used for feedback. The amp's distortion signature is expressing a cubic non-linearity, so mostly a 3rd with succeeding orders falling off rapidly. Since the circuit was quite linear without feedback (the feedback is only used to reduce its gain) the distortion signature is relatively undamaged by the feedback. In comparison to a 2A3 based amp I have on hand, this little amp is better in every way- smoother, lower noise, greater detail, overall things sound more 'real'.  Without feedback it has bandwidth past 100KHz...


This is just my experience, but when typically SETs get compared to PP amps, its never PP amps of the same power and quality. Its also never SET amps compared to PP amps using the same kind of power tubes. In past comparisons I've eliminated these variables; IME when you eliminate those variables and also pay attention to the distortion signature, the advantages of SET are eliminated- you can get overall more pure, smoother and more detailed sound with greater bandwidth and power.

BUT!! for an OTL not to be hindered IN ANY WAY at all by todays "better sounding speakers loads" that have NOT had "their sound compromised" (because they were designed "firstly" too give a very easy load), is a big ask.

There are plenty of them out there in high end audio. The Sound Labs are a good example, although you can’t run them with SETs to do a fair comparison. But usually any speaker that works well with SETs will work well with OTLs too.

I agree that using an auto-former somewhat defeats the purpose of an OTL and would not be an option for me.

There is an enormous difference using an autoformer like the ZERO with OTLs! In most tube amps the output transformer defines the bandwidth of the amp. Because of its very low turns ratio and consequent low inter-winding capacitance, the ZERO has full power bandwidth from 2 Hz to over 1MHz!! -much wider than **any** tube amp. So with an OTL you would have the possibility of the widest bandwidth tube amp (which OTLs are anyway), plus the ability to drive extremely low impedances. We used to offer such a device (called the Z-Music Autoformer) but Paul Speltz started making one too and we felt it easier to sell his than ours.

The one thing that clearly differentiates an SET from any Push-Pull (OTL or not) is the crossover distortion which occurs due to splitting the wave and joining it back. Once I heard and lived with an SET, that distortion was so clearly audible in Push-pull that it was no more acceptable as "right". It is a fundamental compromise.

@pani @charles1dad , sorry, he didn’t nail it; here’s why.


The statement is false. You can easily demonstrate on the bench and in listening that an OTL need not have any crossover distortion whatsoever, at **any** power level. This is one of the traditional arguments for going class A of course (which we’ve been pushing for decades- you don’t get crossover distortion in class A circuits...). The simple fact is there isn’t any crossover distortion in our amps (I can’t speak for other OTLs, but assuming competent design I doubt they have it either).


Probably what you are hearing (because I’m not disputing that you might have heard a difference with a PP amp as opposed to an SET) *is* distortion, just not **crossover** distortion.


This comes from the simple fact that when people compare SETs to PP amps, there are a ton of variables that can cause false conclusions. The idea that its crossover distortion is one of them. So I should point something out: In SETs, the distortion generated is due to a quadratic non-linearity. This results in a fairly substantial 2nd and 3rd harmonic, which masks higher ordered harmonics from the ear (at lower power levels) resulting in smooth sound.


Now if the circuit is fully differential and balanced, you get a different non-linearity known as ’cubic’. This results in the 3rd being most prominent (at a level slightly less than seen in a single-ended circuit). It too masks higher ordered harmonics, also resulting in smooth sound. But the higher orders fall off at a faster rate (due to the cubic function, essentially distortion isn’t compounded as much from stage to stage in the circuit since even orders are cancelled) so the circuit is inherently lower distortion and more transparent, since distortion masks lower level detail.

Now when you **combine** the two non-linearities, as seen in a variety of PP amps with single-ended input, you get algebraic summing of the harmonic orders, resulting in a prominent 5th harmonic. This is well-known; Norman Crowhurst was writing about this 65 years ago. **That** is what you are hearing- not crossover distortion. But if the amp is fully differential you won’t be hearing either one.

In a nutshell, while I don’t doubt that you hear differences, its probably best if you know what those differences are as I’ve described above.

I have used a Graaf Gm20 for about 10 years before upgrading to a Wavac Ec300b some 10 years ago. In short, there is no going back: while the Graaf had twice the stated output power, making that power useful was challenging: using autoformers to ideally match speaker impedance as well as pretty regular rebiasing, neither of which is required by the Wavac. If anything, the Wavac is faster than the Graaf despite its transformers and the bass is substantially more solid. Botj amps are classics in their line of design

What speaker were you using for this comparison? 

The speaker can be an enormous influence when doing comparisons. There are other variables as well- for example most OTLs employ feedback (ours are some of the very few that do not). Some OTLs use pentodes as opposed to triodes. So they sound different; you can't categorically state that because you've heard one OTL you've heard them all and actually be telling the truth.

The distortion signature of the amplifier also plays a big role in the results. In a nutshell there are two types of non-linearities that show up. In single-ended circuits you get a quadratic non-linearity. In a differential circuit you get a cubic non-linearity. The cubic is preferred because its lower distortion; this is because even orders are cancelled and distortion does not compound as much from stage to stage as the signal progresses through the circuit. The 3rd harmonic is the primary distortion product and will mask the presence of the higher orders, allowing for a very smooth sound, but more detailed that that of a single-ended circuit because there is less distortion to mask it.

In a single-ended circuit the primary distortion product is the 2nd and there is usually a prominent 3rd. These two harmonics mask the higher orders so this circuit sounds very smooth as well. The problem is that as the order of the harmonic is increased, its amplitude falls off at a slower rate, causing more low level detail to be masked. But it has a rich sound due to the prodigious 2nd harmonic created.

When you combine the two (such as a single-ended input with a push-pull output), as a good number of OTLs (and other push-pull amps) do, algebraic summing occurs and there usually results more of the 5th harmonic (this has been known a good long time as Norman Crowhurst was writing about this 65 years ago...), which is the main reason SET guys object to the 'sound' of push-pull. But if the PP amp is fully differential you don't have this problem.


Since the feedback used in any tube amplifier is insufficient for it to really do its job, these distortion artifacts remain in the distortion signature.


In a nutshell you have to be really careful about making broad stroke statements; when comparing SETs and OTLs it can go both ways insofar as to which comes out on top. But all OTLs are **faster** than any SET (simply out of the fact that transformers slow down risetimes based on their bandwidth limits; this is simple physics); if it sounds 'slower' its likely because the amp is not matching well with the speaker in the high frequencies.

A single-ended triode amp will always have an output transformer. The transformer is both the boon and doom of the amp- it allows the tube to drive the speaker, but the bigger you make the transformer, the more limited the bandwidth will be.

As a result, the best SETs are the smaller ones (2A3s, 45s, that sort of thing) assuming you have a speaker that is efficient enough to work with them.

OTLs have no output transformer and so don't have the power vs bandwidth issue. Instead you have the opposite problem- its harder to make a practical low-powered OTL because you have to be more careful about choosing a speaker for it.

Usually SETs are zero feedback. OTLs normally have feedback (although ours tend to use little or none). A triode zero-feedback OTL operating class A, like SETs, will generate primarily lower-ordered harmonics, with very little higher-ordered harmonics. This will give either one a relaxed quality.

The ear uses odd-ordered harmonics to tell how loud a sound is, so if those harmonics are not emphasized by the system, the system will have a sense of ease and a lack of hardness.

OTLs are normally push-pull, if built single-ended will produce very little power for the number of tubes involved! Since push-pull operation allows for even-ordered harmonic cancellation, OTLs generally lack some of the romantic lushness of SETs as they will also lack the 2nd harmonic. This is an advantage if you are looking for neutrality and transparency: anytime distortion is present, detail is obscured.

Normally push-pull transformer-coupled amps will have an increase in distortion at very low power, robbing them of that 'inner detail' detail that is part of the 'magic' that SETs are known for. OTLs, like SETs, produce less distortion as power is decreased, giving them a 'magical' quality that they share with SETs; good 'inner detail' and good low level resolution while being very smooth at the same time.

I apologize for the nutshell quality of this post- I'm glossing a lot of things over as this can be a detailed subject!

Aarif, removing distortion never causes an amplifier to have fake neutrality.

OTLs have an economy of scale- the bigger you make them, the more efficient they can run. A big OTL that can make 200 watts will very likely be comfortable on 4 ohms, some smaller ones that make 150 watts might be able to do that also. A 30-watt OTL will not like 4 ohms much at all.

IOW the smaller you make the OTL, the higher you will want the load impedance to be. Many OTL headphone amps are designed for 32 ohms or more, plus they do not have to make more than 0.2 watts. Headphone amplifiers also are a special case because most headphones are designed to operate with amplifiers that have a much higher output impedance than their own impedance.

Without more information about your speakers, it would be hard to make a certain case for either amplifier technology, and we have to include your listening preferences somehow too. Although I have heard very good things about the Graaf, I've not actually heard it, so the following is my opinion only. I think it would have a chance at being a more neutral, musical presentation if the speaker impedance is linear and 7 or 8 ohms at least. I don't agree with some of the tube choices in that amplifier, as I feel that they tend to make the amp sound brighter, so I do have reservations at the same time.

In the end, like so many other things in audio, you will have to audition to know for sure. Remember- its all supposed to be fun :)

The harmonic structure of a good OTL will be the 3rd harmonic, which the only harmonic that the ear considers musical and thus does not object to. SETs have that and the 2nd harmonic.

Paul, what happens with these harmonics is not that they are not audible! It is that they are not **objectionable**. When GE did their experiments in this regard, they found that people will not *object* to nearly 40% of even-ordered content. That is not the same as 'you can't hear it', you certainly can, and audiophiles have terms for the presence of even orders (just like they do for odd orders) in small amounts. From small amounts to larger amounts: Warm, lush, rich, bloom, thick, murky, woolly, muddy. FWIW these terms refer to even ordered harmonics that are in greater supply than the terms that apply to odd ordered harmonics: hard, harsh, brittle, clinical, etc., where the odd-ordered enhancement may well be less than .001%.

With regards to the phase splitter- some OTLs that is an issue, others it is not. In our amps, there is no 'phase splitter' as a stand-alone circuit- that function is integrated into the single gain stage that exists in our amps.

Most SETs have more than one stage of gain- its my contention that if they were able to have only one, they would be a lot closer to the neutrality that our amps routinely express. I've got a 45 SET and a 2A3 SET- they sound great within their dynamic limitations, but they also sound veiled compared even to amps we were making 10 years ago, despite using the best parts available. As far as I have been able to experience, transformers always limit bandwidth, add distortion and rob the music of detail.

Given the example of the speaker above, I would recommend something with more power than the Graaf, or any SET for that matter. The speaker is simply too inefficient for an SET to strut its stuff. You would really need at least 10 more db for that. There are OTLs that can easily drive that load though. IMO 15 watts would not be enough power, unless you listen near-field and at moderate volumes.

Paulfolbrecht, since you asked- in our amps there is only one stage of gain, not two. Most SETs I've seen have either two or three. This is because there is usually some gain in the output, while in our amps there is none.

Crossover distortion, FWIW, is not a function of the 'phase splitter', it is a function of the push-pull operation of the output section. A number of crossover issues stem from output transformers, and some of them stem from class of operation. In a class A push-pull amp, carefully designed, you will not get any crossover distortion at all.

You are right in that if there are crossover issues, they can often appear at low power only, masking low level detail. This is something that simply does not happen in our amps, and it is easy to hear how they have plenty of low level detail, actually more than SETs running at similar power levels do.

You are also right in that harmonic spectra is not the defining issue regarding sound of OTLs and SETs. I should point out here though that most OTLs employ negative feedback, often fairly large amounts, while hardly any SETs use feedback. Our OTLs use little (1 or 2db) or none, so in a way they tend to have more in common with SETs sonically than they they do with a lot of push-pull amps.

Larryi, just for the record, an OTL that has a direct-coupled output does not have to have either a coupling cap or a servo circuit. They can be built to be so stable that the DC Offset can always be very small and only need adjustment occasionally, once every 3-6 months.

The multiple power tube issue is shared by SETs with multiple tubes, Push-pull with multiple tubes and of course transistor amps with multiple output devices. In all cases, it is possible to build the amplifier so the multiple output devices are not a defining characteristic.

Paul, I like the type 45 tube a lot too. Not many speakers that will work with a 45 SET though. I am using a set of Coral Betas, which are fine as long as you don't expect any volume.

Larryi, We've been doing this over 30 years now and all our amps have had a direct coupled output. We've seen many tubes fail in that time. So far, we can still count the number of times that a power tube failure damaged a driver on one hand. In all the cases but one, alternative power tubes had been installed in the amp. So it appears that the amp is quite safe. All of our amps except the big 500-watt unit do not use a servo circuit (we installed one in the big amp, but for convenience only). They are quite stable, and simply don't need a servo. The key is controlling the power tubes, something that you would **think** would be common sense, but it is an issue in a lot of designs. That is one of the reasons why our amps have been so reliable over the years (the other being that the amp does not need negative feedback to work).

The Merlin needs a little more power than most SETs can produce, if you really want to hear what the speaker or the SET is capable of. With that speaker, hands down an OTL will do better.

Martykl, one thing about SETs that makes them quite different from OTLs is the way their distortion signature interacts with the human ear/brain system. SETs tend to have very low distortion at low power, but on transients, where the power occurs, the distortion can be quite a lot higher (10% at full power is quite common).

Thus there is high distortion on transients, and while normally the odd-ordered harmonics are kept at vanishingly low levels, at high power they have an effect. Of course, they are masked by the presence of even-ordered harmonics, but despite that the brain reacts to their presence anyway.

That reaction is a sense that the amplifier somehow is a lot more dynamic and can play louder than it should be able to given its actual power output. You commented about that quality in describing your amplifier above.

OTOH if the amplifier actually has the power to make real spls, and if the distortion is kept low, then the result is that the system is can be more relaxed- that the volume you pick is not based on perceived 'loudness'. IMO, a good system will not have a sense of volume or loudness- it should act that way real music does in that regard.

Marty, I understand what you are saying. I want the amp to play what is good in a recording, and with the poor recordings, not get upset by artifacts in the recordings. In other words, forgiving? -but without giving up anything to resolution. Its a fine line to tread.

Bobby P. has been making certain over the years that Merlins are tube and particularly OTL friendly. OTLs and Merlins are a very common and successful combination!

Hi Marty, most people with VSMs usually go with the M-60. They have enough power to really make the speaker sing.

Everyone says they have the best. In the English language, the meaning of the word requires that there can be only one. All the others are then 'better' but not 'best'.

This has been argued a lot elsewhere on this site. Any consensus on the idea of 'best in my experience'?? instead of merely 'best'?

The Audio Note AN-e LX HE is a fairly benign load and is no problem for an OTL of sufficient power- our customers use our M-60s on that speaker. The speaker definitely benefits from the greater power!

I would say that a well designed SET 2A3/300B  has the warm sound than also well designed an OTL having the same output wattage.

There really aren't any OTLs that make the 2-7 watts as suggested by the above post. OTLs become less practical at lower power levels- driving the real world impedances that one might do with a transformer-coupled amp gets harder. That is one reason OTLs tend to have a lot more power than SETs, and is why our smallest amp makes 30 watts into an 8 ohm load. To make the amp much smaller than that becomes impractical.

However I have found that the OTLs tend to make less of what I have come to call 'loudness cues', IOW higher ordered harmonic distortions (which show up in most SETs when driven over about 20-25% of full power). The result is that the OTL won't sound as loud (even though it is probably louder) and so invites you to use the power it has. This can be done comfortably. The reason SETs sound so dynamic is due to distortion and how that interacts with the ear, not actual volume!