• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Dual SE amplifier? - (not PSE) - Take two!

Sorry Shane, but your thread got messed up somewhere along the line. I find your original suggestion interesting, so hope you don't mind we try again...

There were many good replies untill all the 'my amp is better than your amp' started coming...

Please folks, absolutely no such 'I find SE sound worse than PP and OTL is really the best...etc etc'

Lets keep it to the factual design issues eh?

Ceglar wrote:
Dual Single Ended Amplifier? - (not PSE)

--------------------------------------------------------------------------------

Hi,

Wondering if anyone has seen, built, or heard anything like this..

Two SE output stages per channel, signal at each grid out of phase WRT the other. Two SE output transformers per channel, one of which has its secondary winding reversed so that the waveforms are recombined in phase.

There are some interesting things that supposedly occur in that there is a combined function of the two output transformers.

All of which is outlined in the diagrams and description at;
Dual single ended amplifier by Cohen, Graeme John ? AU1996042023

I'm interested in pretty much anything anyone cares to offer in response.


Thanks,
Shane
-------------------------------------------------------------------------------

The problem with the above is the inverted secondary of one of the SE OPT. The response of a transformer when the 'dots' are connected in phase is not the same as when connected out of phase. Must be b/c OPT are not wound 100% symmetrical, and Pri-Sec leakeage capacitances and inductances are exagerated when the windings swing opposite of each other.
(Kind of like how miller capacitance is actually the static capacitance times the gain(Cgd in FETs or Cgp in tubes - u get the picture)).

How ever, if connecting the outputs in a bridged config, hot positive and hot negative phase, then it should work just fine. Impedances must of course still match, but most OPT have 4,8, and 16ohm taps, so shouldn't be the greatest problem.

Bel Canto made an amp like this some 15years ago: Bel Canto SET 80 monoblock power amplifier | Stereophile.com

I am almost done with a guitar amplifier using two SE OPT b/c I ran out of PP OPT and have tinkering about this idea for years. So this Summer I desided to try it out. Hope to have it finished in a few weekends...

I am building it as a class-AB amp, which means I will let each phase run at about 15watt idle (5881 tubes) and one phase will cut-off while the other goes 'more-on'. That way I will get more power compared to simply running two pure class-A SE in bridged mode.
 
Last edited:
Sorry Shane, but your thread got messed up somewhere along the line. I find your original suggestion interesting, so hope you don't mind we try again...

The problem with the above is the inverted secondary of one of the SE OPT. The response of a transformer when the 'dots' are connected in phase is not the same as when connected out of phase. Must be b/c OPT are not wound 100% symmetrical, and Pri-Sec leakeage capacitances and inductances are exagerated when the windings swing opposite of each other.
(Kind of like how miller capacitance is actually the static capacitance times the gain(Cgd in FETs or Cgp in tubes - u get the picture)).

How ever, if connecting the outputs in a bridged config, hot positive and hot negative phase, then it should work just fine. Impedances must of course still match, but most OPT have 4,8, and 16ohm taps, so shouldn't be the greatest problem.

No problems at all, thank you infact. So you're saying to have the two secondaries in series with each other (the belcanto schem at the link) and not rely on the double phase inversion?.

I think the next step might be to get hands on his Glass Audio Article outlining the idea (Cohen) or a copy of his paper published in 1995 to the AES. AES is probably easier, more clear cut and cost effective.

I'm interested in his take on why he did it the way he did, of course.. and to hold that up to the light of the on-goings here (hopefully).

I'll order the paper and have a read - will get back with details.

Interested in your build progress, along with Tubelabs for the obvious reasons.

Sincerely,
Shane
 
Yes the secondaries in series, like this:
 

Attachments

  • Bridged_SEamp.jpg
    Bridged_SEamp.jpg
    60.5 KB · Views: 793
Interested in your build progress, along with Tubelabs for the obvious reasons.

Some people just like to argue and prove a point. I am not one of them, I would rather play with parts and make stuff.

My lab is rather messed up right now since we had to clean out our rented warehouse that contained 20+ years of collected "stuff" on rather short notice. The building was sold. There is "stuff" stashed in every room of the house.

I only have one PC running right now, so I don't have an FFT box for looking at individual harmonics. It will take a while before everything returns to normal.

In my mind there were two valid discussions in the other thread before it went astray.

One involved driving two SE amp channels out of phase and connecting the OPT secondaries together in series or parallel. Comparisons to be drawn between those two connections, an ordinary P-P OPT, and the SE amp as designed with the two channels paralleled in phase.

The other involved wiring a pair of SE OPT's into a P-P amp since this should emulate the two SE amps.

I had a Tubelab Simple P-P running most of the weekend to listen to the streaming Lollapalooza concert, so I played with some transformers. I immediately discovered the flaw in my thinking that the P-P amp could emulate the pair of SE amps. The Simple P-P is designed to operate in class AB1, usually in pentode mode. The OPT used is 6600 ohms which is 1650 ohms per side. The transition from A to AB occurs at a few watts with this load. Wiring the amp for triode mode and connecting a pair of 3000 ohm SE OPT's will still allow about 2 watts in class A with reduced B+. Pentode mode allowed only a little more power since the load wasn't optimum.

Any attempt to operate the amp above this level resulted in audible distortion. The connection with the two OPT secondaries paralleled worked the best. The series (bridged) connection mentioned in the above schematics was the worst. The problem is occurring because the Simple P-P was not designed to operate in class A and attempting to increase the bias enough for class A required lowering the B+ too much to avoid red plating. The distortion occurs when one tube cuts off and the coupling in the two seperate OPT's isn't good enough to smooth out the "kink" created when the amps output impedance changes. I looked at the output with a scope, but did not hook up any test equipment or make any detailed measurements. The audible results did not warrant it!

So, as discussed in the original posting, we need to stick with class A amps here. The SPP wasn't intended to be a class A amp, so no further experiments are planned using it. I am hesitant to experiment on my two working Simple Single Ended amps since the OPT secondaries are grounded. Modification would require amp dissasembly.

I have a dusty SSE board, plenty of OPT's and several regulated bench power supplies. I will have to hook it up and get it working and preform those experiments. Don't know exactly when though.
 
Interesting Tubelab. In theory though the OPT coupling shouldn't produce more or less of the xover kink...? I am really curious how mine turns out.
Regarding testing the concept with your Simple SE I don't see the problem trying bridging them, since the speaker is connected to each hot output and the secondaries can be grounded. No need to open anything up and mod, just feed the two channels with opposite phases and crank it. (see the attached drawing in my previous post).
 
Interesting Tubelab. In theory though the OPT coupling shouldn't produce more or less of the xover kink...? I am really curious how mine turns out.
Regarding testing the concept with your Simple SE I don't see the problem trying bridging them, since the speaker is connected to each hot output and the secondaries can be grounded. No need to open anything up and mod, just feed the two channels with opposite phases and crank it. (see the attached drawing in my previous post).

Semper,

By your drawing with the OPT's in series there would be a doubling of the output Z, like adding a series resistor to accomodate a 4R speaker on an 8R terminal. So that would have to be figured into the reflected Z. Yes?

20
 
Yes the reflected impedance is halfed.

I just realized something Tubelab has been trying to say. As one tube goes off (as in class-AB) it represents an open circuit to the OPT primary on that 'off-phase'. At that moment the secondary of the other phase, the one going more on, will rather swing the high-Z secondary of the off OPT than the load. So his statement this only works in class-A must be correct. Unless one connects the secondaries in parallel, but then we are back to the problem of one OPT with switched phases...
 
Interesting Tubelab. In theory though the OPT coupling shouldn't produce more or less of the xover kink...?......By your drawing with the OPT's in series there would be a doubling of the output Z

I believe the series connection was the worst case because they were in series, which causes the output Z to be the sum of the output Z for each channel. What is the output Z of a channel when the output tube is cut off? So I have one channel pushing its output into a speaker through a high impedance (and not flat across frequency) network (the OPT connected to a cut off tube. I think the parallel connection could work if the coupling was perfect (which it never will be). It did work better with my admittedly crude setup.

Back when I did the first test (maybe 5 years ago) I also tried taking two SE OPT's apart removing the "I" core pieces, and clamping the "E" pieces (with coils) together. It seemed to work, but again a crude and somewhat uncontrolled experiment.

No need to open anything up and mod, just feed the two channels with opposite phases and crank it.

This is true for the bridged case, and for the true paralleled case, the parallelled out of phase case can not be tested without mod.

I figure that If I am going to do this again, I want to observe every possibility, take some real measurements, and write it all down this time. I also want to try several different types of OPT's and tubes to see how that affects the outcome. I have big and little Edcors, Hammonds 123CSE and 1628SEA and some Transcendars, probably a few more that I don't remember right now. My previous test was only done with the small Edcors. The SSE will fit 6V6GT, 6L6GC, EL34, KT88, and several others with only a twist of the power supply knob.

I have an old SSE board that has been used for experimentation for the past 7 years. It has even been run through the dishwasher once to clean off the goo from an exploded electrolytic. It just keeps on working. I'll set it up once I clean off a spot on the bench and get my second computer running for FFT spectrum plots.
 
Yes the reflected impedance is halfed.

I just realized something Tubelab has been trying to say. As one tube goes off (as in class-AB) it represents an open circuit to the OPT primary on that 'off-phase'. At that moment the secondary of the other phase, the one going more on, will rather swing the high-Z secondary of the off OPT than the load. So his statement this only works in class-A must be correct. Unless one connects the secondaries in parallel, but then we are back to the problem of one OPT with switched phases...

It would be doubled into each primary because they are still independant primaries but each now sees the added secondary in series reflected back.

Also, why run them in AB? Would there ever be a reason to run SE that way unless the current was overloading the PS
 
That equates to 1.5 db difference from the speaker. If getting higher db without clipping amp is the goal, higher efficiency speakers (up to a point) may a better route.

Yep, thats right.

I am building this for guitar btw. The 'typical' amp I seem to build is 15-20watt, but that little extra headroom can be nice in certain settings, and this time I want to build something > 30watts. But my original idea of class-AB (I was even thinking class-B) using two SE OPT was not a good one. Thanks to this thread I don't have to build it to find that out.

There is still one benefit tho, and that is better PSRR thru the balanced topology of a bridged amp. Since guitar amps sound like crap with well filtered and stiff PS, that PSRR is a great advantage. So I will likely build this one as a bridged SE amp, but running class-A, contra PSE. Same output power, better PSRR.
 
There is still one benefit tho, and that is better PSRR thru the balanced topology of a bridged amp. Since guitar amps sound like crap with well filtered and stiff PS, that PSRR is a great advantage. So I will likely build this one as a bridged SE amp, but running class-A, contra PSE. Same output power, better PSRR.

I'm thinking the same with regards to PSRR, also hum rejection with AC heating on the output pair (direct heated tubes) and any bias resistance in the output filament circuit wont add to the Rp.

Thought so far with regard to the filament circuit is to use a common current sink sitting on-top of two pots wired in parallel, wipers feeding the grid returns of the secondary windings of a phase splitting interstage transformer. Would give some means for DC balance, and nice thing about not having a fixed CT in a single OPT (ala PP) is that a centre zero meter can easily be added to balance the currents.

Ordering back issues of the following for some better understanding of at least why the author of the patent link (see original thread), decided it was worthwhile;

Linear Output Stage (Cohen) - Glass Audio Vol.7, #6, 1995
Direct Single Ended Amplifier (Cohen) - Glass Audio Vol.8, #3, 1996

The OPT in his custom offerings were apparently a little different (no iron??) and details should be in the article;

Transmission Line Output Transformers - Glass Audio Vol.7, #5, 1995

Should make for an interesting read.
 
Just another Moderator
Joined 2003
Paid Member
:cop: guys would you like me to merge the technical content from the original thread here? The thread will end up being owned by Ceglar again with his first post in the other thread becoming the first post here. No problem if you would prefer not to. I'm easy either way.

Tony.
 
The Dual Single Ended Article

After quite some time I have managed to locate the original article. Hopefully this can shed some further light on the topology, the article is well written and documented. The distortion comparison is also thoughtful in that one can open one set of curves in a seperate window and they are virtually superimposed over the other, which makes for an easy comparison.

--------------------------------------------------------------------------------

Any comment or futher on topic discussion is very welcome and I have a couple of questions that someone might be able to answer

1) 2H cancellation with 'PP' only occurs when driving an output transformer?
2) If a SE to PP IT was used between a single driver and output(s), would you expect the typical SE distortion from that transformer loaded driver, even with zero grid current in the secondary windings?
3) Any issues with using a single bias resistor (unbypassed) that is common to both output stages?

--------------------------------------------------------------------------------

The excerpts here have been presented with kimd permission of the entity that holds the copyrights to the published article which appeared in Glass Audio Magazine Vol. 8 No. 3, 1996. Thanks to Mr. Barraclough and AA/CC/Elektor USA

Supplements to this article might include;
Transmission Line Output Transformers - Glass Audio Vol. 7 No. 5, 1995
Linear Output Stages - Glass Audio Vol. 7 No. 7, 1995



THE DUAL SINGLE ENDED AMPLIFIER

"The revival of the SE ampflifier is becoming more appreciated. By combining SE ampflifiers in the correct way, and driven correctly a new amplifier has been developed. Distortion figures lower than both the SE and PP types have been obtained. Listening tests have confirmed the sonic attributes of the Dual Single Ended amplifier."


"A revival of the operation of a single output device, in particular the triode valve, has emerged again for serious audiophile listening. The second harmonic distortion that can be present with SE operation is either desired, accepted or ignored in the quest for high quality amplifiers. Many engineers disregard valve amplifiers, particularly the SE type".

"By using two SE amplifiers per channel, driven correctly and combined in a such a way that the total power is added while reducing the harmonic distortion, a new SE amplifier techique has been developed.

Two identical SE output stages with their own output transformers are driven in a differential manner. The combined output is added, in series or parallel, at the secondaries of the output transformer.

Each output stage operates in Class A with the normal high standing current and high flux in each transformer in true SE operation. As the flux density is increased in one output transformer, it is reduced in the other transformer, with both the second and third harmonic distortion products being reduced".

Push-Pull

"High third order distortion can be present. Symetrical saturation reduces second order products but not necessarily third order products.

Any unbalance of the PP primary currents can cause high levels of distortion (Ref. 1). Typically, ten percent unbalance of DC current can cause several percent harmonic distortion due to change of flux density (Fig. 10).

Large changes of inductance and permeability are present due to the operating flux varying from zero to maximum each half cycle (Fig. 8, Ref. 2,3 & 4). Changes of up to 5 to 1 are typical with a flux density peak followed by a reduction of flux at high levels. The effect is compounded at the speaker resonance frequencies."

Single Ended

"The high standing current requires a large output transformer with an air gap. This results in low transformer core distortion and minimum inductance changes with changing flux density. The flux lines are in one direction only as they do not go to zero over any portion of the Class A operation (Fig. 9).

Valves are usually used today in SE operation, and due to their low output impedance the transformer distortion is less. If the triode is operated in a linear manner with low distortion, then the transformer distortion dominates and this can be minimised with good design.

For a low distortion valve stage, the transformer distortion dominates due to flux saturation at high levels. At low frequencies the change in inductance with flux density also modifies the affect of the above conditions."

Output Valves

"when a triode valve is used, the transformer is droven from a low impedance, which produces a linear voltage across the transformer, with the transformer current having the major distortion component.

If a Pentode or Tetrode valve is employed, the transformer is driven from a high impedance producing a linear current through the transformer, and the major distortion is the voltage across the transformer (Ref.1, as above).

A Pentode or Tetrode valve connected as a partial triodes (sometimes referred to as Ultra Linear, or Distributed Loading_, has the screen grid of the valcve connected to a tap on the transformer primary winding. The relative high efficiency of these valves is retained with some of the benefits of triode valves being obtained. This confuguration has a medium output impedance and the transformer distortion is a combination of voltage and current. This can result in a lower distortion power match to the loudspeaker load.

If a low amount, or zero overall feedback is desired, then triodes need to be used to obtain an inherently low output impedance, and hence an adequate damping factor. For zero feedback operation, inherently linear output stages using triode valves and wide bandwidth low distortion transformers need to be employed (Ref.5 & 6)."

Curvature Distortion

"If Class A PP is considered, it may be assumed to be the circuit topology which gives the lowest distortion. The major benefit of PP is to reduce the second order distortion for a given power output per anode dissipation, and compared to a SE stage for the same dissipation this may be the case.

However, Class A PP can have considerable third order (third and fifth harmonic etc) due to each half cycle rounding, due to either the output valve or the PP transformer. That is, each half cycle, can have severe rounding which can cancel for second harmonics, but can have a severe third harmonic distortion component.

The SE amplifier can have both second and third order components.

However, when two SE output stages are correctly added, each half cycle has cancellation of both second AND third order components, that is, the rounding of one stage due to and increase of both valve and transformer currents tends to be cancelled by the decrease of current in the other stage. As both stages are identical, but operating in the opposite phase, its is possible that all forms of distortion are reduced during each half cycle, and that the second harmonic distortion is not dependant on the next half cycle.

The Dual Single Ended concept therefore tends to reduce all distortion components during each half cycle for both valve and transformer distortion

In Figures 11, 12 & 13 are shown transformer distortion due to flux changes (B) with magnetizing current (H) for PP, SE and Dual SE respectively.

The PP case, (Fig. 11), shows a high level of third harmonic distortion (n.b. a square wave has low second but high third harmonic distortion).

The SE case (Fig. 12), has a considerable level of both second and third harmonic distortion but has other attributes

The Dual SE amplifier concept, (Fig. 13), using two SE stages driven out of phase bt added to give maximum power, tends to cancel distortion and this reduces both second and third harmonic distortion. One Single Ended stage curve is drawn updside down to show the ideal resultant when combined.

Demonstration

"To show the difference is distortion products between PP, SE and Dual SE output stage, a low distortion amplifier was developed. It was designed to have a high output impedance in contrast to the normal desirable low output impedance amplifiers. The reason is to achieve a linear current through the transformer primary winding so that the primary voltage distortion could be readily observed on an oscilloscope. The general form of this test ampflifier is shown in Figure 14.

The output valves had high value cathode resistors with feedback from their cathodes back to the input stage. A linear current drive to the output transformer resulted, with no feedback from the transformer itself.

A Trimax transformer TA1044 was used for the PP tests and two SE transformers Trimax TA851A were used for the Dual SE tests. The two SE transformer secondaries were combines to give maximum output power. The amplifier was driven in a balanced differential manner whith both of its halves being independant, identical SE designs.

The results are shown in Photographs 1 to 6 for a 20Hz input signal. The outputs were taken to an optimum resistive load. The total output power was 5 watts which was less than half of the transformers' rated power and well inside the linear operation of the current output Class A valves. The anode voltage waveforms were taken by inverting one channel of a two channel oscilloscope."

"In Photograph 4 is shown the Dual SE anode waveforms across the two individual SE transformers. Again, one oscilloscope channel is inverted. The typical SE rounding is seen on each waveform. However, when the two anode voltage waveforms are added in the oscilloscope, the two seperate SE wave-forms are combined to show the Dual SE waveform in Photograph 5. Note the absence of the kink as in the PP case. In Photograph 6 is the combined secondary of the two SE transformers. The waveforms visually appear better than the PP case without the kind present, also the wave analysers measurements show less distortion overall and in particular less third harmonic distortion

The above tests may not represent all cases of comparison but do show the mechanism for what might be considered high distortion in each SE output, and in fact has less distortion than a similar power PP case, when the two independent SE stages are combined in the Dual SE configuration".


---------------------
Regards,
Shane
 

Attachments

  • img001.jpg
    img001.jpg
    148.1 KB · Views: 349
  • img002.jpg
    img002.jpg
    245.9 KB · Views: 339
  • img003.jpg
    img003.jpg
    137.8 KB · Views: 335
  • img004.jpg
    img004.jpg
    201.3 KB · Views: 337
  • img005.jpg
    img005.jpg
    171.6 KB · Views: 319
  • img006.jpg
    img006.jpg
    172.8 KB · Views: 146
Last edited: