My take on a Single Ended amp

Kevin,
I apologize if I divert the discussion a tiny bit, but I would like to ask Dowto to expand on his statement about PSU.

Dowto100, you've stated on several occasions (in this and other threads) that PSU is critical for SE amp, and I fully agree with you on that. But what does it mean (for you) to build a "high-quality" PSU? Can you list a "full" set of specifications? And how do you determine/measure the performance of your PSU as it relates to the amp performance? Are these measurement objective and quantitative (i.e. done with the scope/analyzer and can be plugged in to Excel for analysis or graphing), or are they subjective and based on the listening tests? BTW, I either way is fine with me, please don't take my questions as insult or challenge, but rather as a learning/sharing inquiry.

Sincerely, Paul.


Hi,

Good question. Personal Background.

My first audio mentor was Robert W. Fulton, of Fulton Musical Industries . He was responsible for the FMI-80, FMI J-Modular, Premiere speakers ( 12HZ to 100 kHZ ) . I first met RWF in 1978, and in 1981, I built my first DHT amp, 2A3 P-P, from an AudioNote Kit for Audio and LP Reviewer Neil G. Levinson. In 1982, I decided to do my own design. The very first thing RWF told me was " to make all the chokes to the 2A3 Finals 20 Ohms or less, IF I can find them. " He also suggested two L/C sections in series, as being superior. A six chassis 700 pound monster evolved, 1982-1986 .....P-P-P 6B4Gs.

Sadly, RWF died in 1988, and pre-internet, I corresponded with other amateur audio designers and DIYers, People like Art Loesch, Herb Reichert, JC Morrison, Jim Dowdy, and Dennis Fraker. I shared freely with them " 20 Ohms or less " on Ls.

Reichert's Flesh and Blood 300B amp had my "only $8.00 surplus - priced potted 3 HY at 24 Ohms DCR choke ", as part of his final design in that Sound Practices magazine article.

Dennis Fraker introduced me to a PhD EE - audiophile, who was head of the EE Department at the University of Alabama. Dr. Charles E. Halijac gave me a scholarly run down on DCRs in power supplies, and how conventional EE supply designs " by the book " were " outmoded and out-to-lunch ", something I already sorta knew, and it alluded to RWF's " 20 Ohms or less ".

I think, in around 1998, Dennus Fraker, who was my brightest and best audio friend, and a " second audio mentor", developed his first Serious Stereo DC 2A3 SE audio amp. It used two 10 Ohm C-2708 Stancors, and two L/Cs, ....right up my alley. But Dennis' implementation was FAR FAR advanced, from just using Low DCR Chokes. He also BEAUTIFULLY devised a scheme where the chokes and the amount of capacitance, in uFs, was also low. This was HIS own LSES supply, totally unique in the history of audio. LSES stands for Low Stored Energy Supply.

From Dennis' second year showing at RMAF / Denver, 2006, I attended those shows annually, to hear his SE 2A3 amp implementations, and anything else that sounded good to me, in audio. Fun. I applied his approach to my SE amps, and enjoyed unique results. When I A-Bed my SE 45 amp, with LSES, against, say, a local audio friend's Yamomoto A-08, my implementation put that Yamamoto SE 45 amp to TOTAL shame !! We both heard it instantly, we KNEW, and we were encouraged.

I eventually wanted to share this supply , based on a MONO 45 or 2A3 amp, with good people on the Audio Asylum. There, I had mixed results. The " 100 year old textbook theory guys, " ALL of them never EVER hearing Dennis' amps at a RMAF show, wanted to criticise me for parting with their beloved theory of " critical inductance ". Two very very sharp EEs, Messers John L. Hasquin, and John Swenson, embraced the design concept, developed solid engineering theory to explain why it was so good, and developed their OWN version of LSES, called by Swenson, a FLYWHEEL. These two actually built a circuit and LISTENED to what they built...amazing !! If anyone wants to read their chronological AA Posts, I have them saved and can email them to you.

I did try Swenson's Flywheel , ( which used Fraker's design concepts in different way ), and eventually, I personally went back to pure LSES, ala Dennis, - upon HEARING his RMAF exhibits annually, and have stayed there, and its been developed additionally, in subtle ways.

So, ALL OF THE ABOVE, represents 37 years of my DIY audio life, dedicated to this subject ( 2018-1981) .

In a separate post, I will address your questions, on HOW I would employ LSES in 2018, and what constitutes a good SE amp power supply, in my experience. This post, addressed my personal background for folks to comprehend.

Have fun, I always do.

'Am presently prototyping a SE two stage Direct Coupled 6AQ5 amp, ultralinear. That tube itself, deserves a great circuit implementation, IMHO.
 

Attachments

  • TMA Color Photos , one channel top, both in VW 1 of 4.pdf
    314.3 KB · Views: 106
Last edited:
My one qualm with the LSES idea, is that by using a low-value choke and low-value cap- you're creating a resonant circuit that's likely got a corner frequency within the audio band.

I've heard tube amp power supplies with this issue- and my usual complaint, is reduced low bass energy and impact- and what bass is there, tended to be tubby and ill-defined. I've modified such designs, with additional capacitance and/or inductance- getting the resonant frequency below the audio band, and reducing the Q of the resonant circuit to a low enough level to not overshoot- and the bass has almost always improved.

Incidentally- I HAVE heard Dennis Fraker's designs- at RMAF 2017. I actually directly compared his room, with another room, also using a monitor speaker based on the Altec 604, among other rooms I sought out, with tube amplification, anywhere I could find them. I thought Dennis' system (with the SE amps and Altec 604s) didn't have nearly the bass weight or definition of the other 604s, in fact. The sound in Dennis' room was sort of "thin", by comparison, to my ears.

I didn't even know about his power supply designs at that time- but if they're designed in the way described above, the sound I heard actually makes more sense to me...

Of course, I will admit- I'm more attuned and drawn to the bass authority, reach and clarity, generated by Kevin Hayes' VAC amps. It's hard for ANYTHING else to compete with that, IMHO... and that includes all other amp designs of all types, tube or solid state, that I've encountered, period...

Regards,
Gordon.
 
I am slowly making progress on this thing. I've got a jig built that will allow me to prototype the basic topology of the amplifier and let me play with biasing, B+ voltage, and resistor value choices. It'll be slow going for a while though, since we gotta get our daughter off to college in another few weeks.

Can't do much else on honing in on the circuit until I order the output transformers. That will take, what, 6 weeks once I place the order from Edcor....:(

Here's the jig:

IMG_3194.JPG
 
I am not sure you can drive the KT88 grid load with a 12AT7. I designed a similar SE amp using orig. 6550s but used a 6DJ8 into 47 k ohms.
It has an output Z of 6k ohms, x10 for bridging comes out to 60 k Ohms. A bit of a mismatch but it works well. BTW, it produces 7 w RMS and about 12w peak, which drives my 90 dB speakers well enough unless a huge bass signal comes through. Then it just goes SPLAT for a second.
 
Paul they will be the CXSE-25-8-5K.

Triode, what is the difficult part of a 12AT7 in driving a KT88? Is it voltage swing, or maybe recharging the coupling cap quick enough without causing blocking distortion? This splat you describe sounds like blocking distortion? I might have to resort to buffering. I'll find out soon enough when I prototype it.
 
I am not sure you can drive the KT88 grid load with a 12AT7. I designed a similar SE amp using orig. 6550s but used a 6DJ8 into 47 k ohms.
It has an output Z of 6k ohms, x10 for bridging comes out to 60 k Ohms. A bit of a mismatch but it works well. BTW, it produces 7 w RMS and about 12w peak, which drives my 90 dB speakers well enough unless a huge bass signal comes through. Then it just goes SPLAT for a second.
In an RH type SEP amp, where fb is from the plate of the power tube to the plate of driver tube. The 12at7 has no problem driving the kt88. So, while Kward's circuit is not exactly that it is similar but no Schade FB. I have made many RH type amps with the 12at7, and also others with different driver tubes like 6sn7 and 6sl7.
 
The Tubelab Simple SE amp uses a 12AT7 to drive KT88 finals amongst others. I don't know if it does so optimally but folks seem to like it.
 
Thanks. I'm thinking it should be okay since 12AT7s are used routinely to drive output stages in power amps (mostly guitar I think, but some hifi as well). I am concerned about the size of the coupling cap as this could increase the propensity for blocking distortion. I might resort to partial cathode bias so that I can use a larger grid leak resistor which would allow me to use a smaller coupling cap.
 
Finally got the Edcor CXSE-25-8-5K outputs delivered. These are big heavy lugs of iron. Size and weight-wise, they look like about the size of Dynaco A431 iron (something I'm familiar with).

IMG_3200.JPG

Measuring them on the bench:

Sample 1:

Unloaded: 4.2K primary
Loaded (8.0 ohm non inductive load): 4.7K primary
Screen tap: 39.4% of voltage at plate
Frequency: 60 Hz

Sample 2:
Unloaded: 4.3K primary
Loaded (8.0 ohm non inductive load): 4.8K primary
Screen tap: 40.0% of voltage at plate
Frequency 60 Hz.

So...screen taps are right on the money, but these samples are not quite 5K primaries as advertised. Nonetheless, close enough I think, where I can work with them. The load line will be slightly steeper than I anticipated so I may need to adjust the desired plate voltage a smidge to get equal swing on either side of the quiescent point.

Anyway, next step is to mock up one channel and see if I can determine the optimum plate voltage to use. Once I know that I can hone in on a workable power transformer.
 
Last edited:
I have a day off today (Labor day) so I get to work on some hobbies! I've mocked up a single channel of this amp. Here's a picture of the rat's nest:

IMG_3201.JPG

Basically what I have is 1/2 of a 12AT7 for gain stage, capacitor coupled to a Tung Sol KT120 output tube. I decided to go with the KT120's because I have a quad of them I recently purchased. I don't have any KT88's in house, and I don't want to fork out the money for a nice pair at the moment, so I think I will make this a KT120-based amp instead. Using the bench power supply, I then adjusted plate and negative bias voltages so as to obtain the lowest distortion reading measured on the output across an 8 ohm resistor load. Here's what I got:

  • B+ voltage: 450V
  • Bias voltage: -49.6V
  • Cathode current at idle: 90 mA.
  • Screen current at idle: 10.3 mA
  • Plate dissipation at idle: 36 watts (60% of design center max dissipation for a KT120 tube. If using a KT88, this would be approx 85%)
  • Power output into a dummy 8 ohm load: 14 watts before significant wave form rounding on the top sides of the waves (but it's only a single channel driven so far. I expect it will drop to maybe 12 watts when both channels are driven)
  • Distortion at 14 watts output at 1KHz: 1.5% THD (but no feedback is applied yet).

I'm noticing the output waveform clips cleanly on the bottom side of the wave form, but the top side rolls the sine wave rounder and does not ever produce a sharp clipping corner until way way into overdriving the output stage. Not ever having built or even worked on an SE amp before, it was a little disconcerting initially to see this behavior--not like a PP amp at all. But on the other hand, I think this is expected behavior given this kind of topology in a SE amp--tube behaves differently when it's forced into cutoff on the down side vs. forcing into grid current on the upside.

Anyway, the next thing to do is add some feedback and take a look at the HF performance of these Edcor output transformers....
 
Last edited:
Specs on this particular transformer say max power handling capacity is 25 watts and max current through the primary is 200 mA.

The DC (quiescent) plate power dissipation at idle is 36 watts as quoted above. That's calculated as P = I * E, where E is the voltage drop across the tube (450V DC) multiplied by the current through the plate (about 80 mA DC). So the current going through the primary at idle is 80 mA through plate + 10 mA through the screen = 90 mA total.

The AC (RMS) power generated at max output power is calculated as P = I * E as above, but I and E are the RMS values of the voltage and current swings which generate that max power output. Peak to Peak current swing at max power output is approx 160 mA (got that from the load line I drew), or 56 mA RMS. Peak to peak voltage swing at max power is about 850V (again from the load line), or 300V RMS. So AC plate power generated is 300 V * 56 mA = 16.84 watts.

The transformers are rated for 25 watts but only 16.84 watts are consumed. The transformers are rated for 200 mA current through the primary, but only 80 mA DC + 56 mA RMS = 136 mA total is consumed. So yes I do believe these output transformers can handle both the power and current demands of this amp simultaneously, and actually with a good bit of headroom to spare. This was the reason I picked these particular output transformers so that I would not be bandwidth limited even at max power the amp will produce.
 
Last edited:
I have a day off today (Labor day) so I get to work on some hobbies! I've mocked up a single channel of this amp. Here's a picture of the rat's nest:

View attachment 1272290

Basically what I have is 1/2 of a 12AT7 for gain stage, capacitor coupled to a Tung Sol KT120 output tube. I decided to go with the KT120's because I have a quad of them I recently purchased. I don't have any KT88's in house, and I don't want to fork out the money for a nice pair at the moment, so I think I will make this a KT120-based amp instead. Using the bench power supply, I then adjusted plate and negative bias voltages so as to obtain the lowest distortion reading measured on the output across an 8 ohm resistor load. Here's what I got:

  • B+ voltage: 450V
  • Bias voltage: -49.6V
  • Cathode current at idle: 90 mA.
  • Screen current at idle: 10.3 mA
  • Plate dissipation at idle: 36 watts (60% of design center max dissipation for a KT120 tube. If using a KT88, this would be approx 85%)
  • Power output into a dummy 8 ohm load: 14 watts before significant wave form rounding on the top sides of the waves (but it's only a single channel driven so far. I expect it will drop to maybe 12 watts when both channels are driven)
  • Distortion at 14 watts output at 1KHz: 1.5% THD (but no feedback is applied yet).

I'm noticing the output waveform clips cleanly on the bottom side of the wave form, but the top side rolls the sine wave rounder and does not ever produce a sharp clipping corner until way way into overdriving the output stage. Not ever having built or even worked on an SE amp before, it was a little disconcerting initially to see this behavior--not like a PP amp at all. But on the other hand, I think this is expected behavior given this kind of topology in a SE amp--tube behaves differently when it's forced into cutoff on the down side vs. forcing into grid current on the upside.

Anyway, the next thing to do is add some feedback and take a look at the HF performance of these Edcor output transformers....
Hi Kev, did you give a listen to the circuit? In my SEP KT120 amp i am using 6sn7 for the driver but the original design was for a 12at7. The circuit calls for a 100K feedback resistor from the plate of the power tube to the plate of the driver tube. This amp also uses fixed bias and i am running the cathode current at 110 ma . The Edcor i use is the same size as yours but 3.5K primary impedance if i am remembering correctly. I did an impromptu amp shootout with a SEP 7591 amp from a "Airline" console. This amp is the stock circuit of 12ax7 driving the 7591 with a beefed up power supply. I was really surprised how much better the big RH amp sounds with plate to plate fb. The airline amp has global NFB. After making many SEP i am pretty much convinced that Schade type FB is better sounding than global NFB. I guess with your current test bed it would be interesting if you could compare the sound quality between the 2 different FB circuits.
 
So...

With the application of 10 dB of global feedback encompassing the output transformer, I have the following situation:
  • At 14 watts output power (still single channel driven), THD is 0.52% at 1KHz. This is actually a bit lower than I was expecting, but might rise a bit when the power supply drives both channels simultaneously. I'm driving the amp right now with my bench PS, which regulates both high voltage output and negative bias voltage output. So this is the best I think I can expect.
  • Input sensitivity is 1.63V. This is a bit more sensitive than I was expecting. Apparently I have mis-calculated the closed loop gain of the 12AT7. I'll have to go back and measure this when I get more time.

As mentioned, the only thing added in the FB loop is the feedback resistor itself (no phase adjust cap or any other HF compensation yet). Here's what an 8 KHz square wave at 1 watt output into a purely resistive load looks like so far.

IMG_3206.JPG

This seems extremely promising. The tops of the square waves are pretty flat after the initial ring settles. There appears to be very little winding resonance up in the higher octaves above 20 KHz and this is a very good sign. It should be quite straight forward to add some HF compensation now to tame that initial overshoot.

So things are looking on the up and up!

Next thing to do will be to add some HF compensation, and then validate HF and LF stability. If I get that far and things still look good, I think I will have validated this design enough for me to take courage to start building the actual amp.

But a nagging feeling persists that something could rear its ugly head, so I am not claiming victory just yet...
 
Last edited:
Primo--no I haven't listened to the amp at all yet. I want to try that Schade style feedback at some point. Maybe I could use a 6FQ7 in place of a 6SN7 on the test jig. I'm stoked right now though that the Edcor OPTs seem to be really well wound. Not at all like my Edcor PP output transformer experiences...
 
The transformers are rated for 200 mA current through the primary, but only 80 mA DC + 56 mA RMS = 136 mA total is consumed.
And I made a typo. It should be 90 mA DC + 56 mA RMS = 146 mA total. (accounting for screen current).

But now I wonder how the true power through the primary is figured since the idle current on the plate is 80ma, through the primary, it seems that would be the swing point +/- the AC current, not 0ma as a normal RMS would be figured from.
Yes exactly, and this was accounted for in my calculations. 90 mA DC added to 56 mA AC (RMS). In other words I started with the static (quiescent) current through the primary and added to it the AC current seen from running a signal through it at maximum signal excursion (max power).

One part that may be confusing here is how much DC power is consumed or lost in the transformer due to the static (quiescent) idle current on the primary. It is a little bit for sure, but I ignored it in my calculations since it is pretty small. Because we're talking about DC current in this case, power loss would be caused by the primary wire resistance. The spec sheet for this transformer says it's 85 ohms. So DC power loss through the primary due to wire resistance would be 90 mA * 90 mA * 85 ohms = 0.68 watt.

And I guess to be complete in the discussion, there's also some power loss through the wire resistance from the AC signal. It would be 56 mA * 56 mA * 85 ohms = 0.26 watt.

So in total, 0.68 watt is lost inside the transformer due to wire resistance at idle, and 0.68 watt + 0.26 watt = 0.94 watt is lost inside the transformer due to wire resistance of the primary at full power output.
 
Last edited:
Referring to the schematic below, which is what is built on the test jig currently:

upload_2018-9-4_23-2-0.png

10 dB of feedback is obtained with a 330 ohm feedback resistor into a 62 ohm lower cathode resistor. Surprisingly, the output transformer is relatively immune to the size of the feedback cap. I set it at 2200 pF and it tamed the initial overshoot on the square wave test quite a bit, but not quite enough for my tastes. Larger feedback caps tended to round the initial edge of the square wave too much while not reducing overshoot any more than the 2200 pF cap did. So I then added a 220 pF screen feedback cap and that got it about where I wanted it.

With the HF tuning parameters shown in the above schematic, the amp is rock stable at LF--settling occurs within 1/2 cycle under no load conditions, and is so quick under an 8 ohm load I can't really measure it. HF stability is stable under no load and with any size cap load down to 0.01 uF (smallest size tested). I can see some weird behavior when I drive the amp with a 1.2 MHz sine signal. Nonetheless I can't say it's ultimately stable since there does appear to be some oscillation at 1.2 MHz but it is way way down in amplitude. I suspect it's caused from the maze of connecting wires and no metal chassis to keep out the interference.

Relative to the amplitude at 1 watt output at 1KHz, frequency response is flat from 20 Hz to 34 KHz. Frequency response rises to + 0.25 dB at about 12 Hz and then trails off again at lower frequencies. Frequency response is -1 dB down at 46 KHz. There are some normal peaks and valleys in frequency response from 46 KHz to about 100 KHz as I would expect but they are way down in amplitude at that high of frequency. I expect these peaks and valleys are due to winding resonance.

So....this seems to me to be not too bad for a first attempt. This just being an exercise to determine if I continue with the design or if I need to rethink something. Results are promising, and it says to me to continue with the design. So that's that.

Here's the same 8 Khz square wave response with HF tuning parameters as shown in the above schematic into an 8 ohm resistive only load:

IMG_3208.JPG

I think what I want to do next is measure power bandwidth at 14 watts output, and then also measure THD across the audio band at 14 watts output. If those numbers report favorably, I think I will be ready to build the circuit in the amp chassis.

I wish I could listen to the amp in current form, not that I can tell a whole lot with a single channel anyway, but I'm thinking it's too much of a rat's nest to cart it to the listening room, and besides I need the bench supply to drive it currently.

The rat's nest:
IMG_3210.JPG
 
Last edited:
Power bandwidth and distortion sweep measurements.

Well, it ain't no PP amp in the performance department.

Power bandwidth @ 14 watts output, single channel driven
40 Hz - 20 KHz +0/-0 dB
40 Hz - 39 KHz +0/-1 dB

Using my HP 339A Distortion Set analyzer, THD distortion sweep measurements at 14 watts output, single channel driven are the following:
30 Hz: 6.2%
40 Hz: 1.65%
50 Hz: 1.15%
60 Hz: 0.93%
100 Hz: 0.64%
1 KHz: 0.47%
5 KHz: 0.51%
10 KHz: 0.58%
15 KHz: 0.75%
20 KHz: 1.0%

Note: At 1 KHz, the amp will deliver 19 watts output (single channel driven) with < 0.5% distortion. It can't do it consistently across the bandwidth though.

So....

Pretty clear this design supports a power bandwidth of 40 Hz - 20 KHz at 14 watts output. Visually the sine wave starts to fall apart (looks more like a stepped ski jump) at 39 Hz and below, at 14 watts output. It fares a little better in the high end, as good as a semi-decent Push Pull amp here.

Well, I'm not sure I could have expected better bottom end performance. I knew it wouldn't be great (as compared to a PP amp) but I was hoping for under 1% distortion down to 30 Hz, like I think I could get without too much trouble in a PP amp. That isn't reality though for these SE transformers. So it is.

Still the good news is I don't have wild out of control distortion in the upper octaves. That's a great sign. So onward we go...

Next step: I think I want to try just one more thing...that is swapping positions of the 12AT7 680Ω cathode resistor and associated cathode bypass cap and 68Ω cathode resistors. That will change the series/parallel orientation of the feedback cap relative to the cathode bypass cap. Not sure what this will bring, but it's dead easy to test on the test jig, so why not.

I also think I will redo the distortion and sweep measurements tomorrow night when I'm a little more coherent just to validate these results.

I'm also starting to draw up the final build schematic and I think I can have that posted within a week or so. (still waiting for a show stopper to rear it's head...I'm not seeing it...so maybe I'm on the down hill side of design validation now.)
 
Last edited:
Hi Kev. the test results look decent. Could you check to see what some Schade FB would do in terms of your tests?. In an RH type amp it would simply be a 100k 1w resistor between the plate of the kt120 and the plate of the at7. You would need to disconnect the fb connection at the 8 ohm tap. cheers, primo
 
Unloaded in this design, the 12AT7 has a 36K plate load. Driving the KT120 into the 47K grid leak resistor reduces the effective plate load on the 12AT7 stage down to 36K || 47K. Now on top of that, add plate to plate feedback with say a 100K resistor. This adds another effective load resistor of 100K divided by the gain of output stage. If output stage has gain of 7, effective load resistor is 14K. So total effective load the 12AT7 sees at AC is 36K || 47K || 14K = 8.3K. That's a hugely heavy load demand on the 12AT7 at the bias point I've chosen.

Graphically, the 12AT7 load line rotates clockwise around the quiescent point when you reduce its effective load. So the load line becomes significantly more vertical. Effective gain is lost (thus less gain for feedback to work), and lower input sensitivity is a result. Trying to keep input sensitivity in the 2V to 2.5V range will limit the amount of feedback that can be applied. If I make the grid leak resistor on the output stage 100K instead of 47K (violating max grid leak resistor size for the KT120 in fixed bias config), it increases effective load to 9.3K, which is a little better, but still hard to drive. If I swap over to cathode bias, I can make the grid leak resistor easily 100K, maybe 150K, which will help a little more, but now we're not using the design that was worked up for this thread. It's a whole new amp now.

I want to be able to drive the KT88/KT120 cleanly w/o introducing distortion, while still having enough gain for feedback to be effective, while still having decent input sensitivity, while still having low distortion drive, while still having enough current drive as to not be too sluggish in driving the R/C coupling to the output stage. I may be able to decrease the coupling cap size a bit, say to 0.22 uF, and thread the needle on a certain bias point that will achieve all those other requirements simultaneously.... But I fear if I just slapped on some Schade feedback while changing nothing else, I would have worse performance.

If a direct coupled buffer were added between the 12AT7 and KT120, that would make the situation much better, but now there's another stage needed, so a more complicated design. Shoot, if I'm going to have three stages, I might as well build a Mullard-style push pull amp circuit.

But let me think on it some more. I want to draw some AC load lines for this situation and see if I can find a design that would work. I may be making a 'mountain out of a molehill' so to speak.
 
Last edited:
Back
Top Bottom