My take on a Single Ended amp

[kward]

It's crazy, I know...I'm finally biting the bullet and building a single ended amp. All the fuss you single ended advocates have been making over the years has piqued my interest and I'm going to try it!

I'd appreciate any feedback or thoughts anyone may have on this design.

I have only ~90 dB sensitive speakers, so I need something with a little power behind it. A flea watt type won't work at all for me. Minimum power I can live with is maybe 10 watts or so in my small-ish listening room.

Here's the schematic I worked up.



First thing you will notice is it's not cathode biased. I specified fixed bias only because I didn't want the heat dissipation of a big ol' cathode resistor under the chassis, plus it will give maybe a watt or so more output power (not that I will be able to notice...).

I've heard very good things about these Edcor transformers. The CX line is the biggest baddest SE OPTs they make. So I'll try them...

I'm not a fan of directly coupling the driver to the KT88 because that means either 1) pushing up the cathode voltage of the KT88 so high that you need a big electrolytic to bypass it, and I just can't imagine that a big electrolytic will sound sonically better than a small film coupling cap, or 2) shifting down the driver plate and cathode voltages with a negative supply rail so you can direct couple via a fixed bias approach, but that then requires capacitor coupling the input. So if I need cap coupling anyway, might as well place it in between driver and output because that is the simpler solution, and requires only a very small current negative bias supply.

Ideally I only wanted one stage in front of the power tube, but driving a KT88 with only one stage in front of it is a bit of a challenge. The only tube I know of that can drive a KT88 with the gain needed, and that is readily available in new production is a 12AT7. But the other thought I had here was to use a 12AX7 and direct couple it to a buffer (cathode follower), which then drives the KT88 (cap coupled). I know that would work well, and ultimately I may end up going that way, but it adds another stage. The buffer could be made from a 6FQ7 or any tube of that ilk. One reason I didn't go that way for now is I wanted it as simple as possible. I could use a mosfet driver, but that's again getting too complicated, plus I'm not sure I want sand in the signal path. (one day I will probably try it though).

I think a bit of global feedback is good (I know I am blaspheming against the true SE crowd). With 7.4 dB feedback, the math says input sensitivity is 2V rms. That's a good place to be. But I would like to add a little more feedback (say 10 to 12 dB), but the gain of the 12AT7 doesn't support it without raising the input sensitivity. I'd like to keep sensitivity at 2V rms to work with a wider variety of preamps, but ultimately the higher level of feedback to get a better damping factor may win out over a single driver stage design.

I am aware of a type of feedback taken from the OPT primary (either from the plate or from the UL tap) and feeding it back to the driver. I will probably eventually experiment with those configurations. But for now, I just need to get a simple base line amp built so I can get some experience here.

The KT88 idles at 80% of max plate dissipation. The math says with both channels driven, voltage sags a bit to 430V, and power output is then about 12 - 13 watts at the speaker terminals (that's assuming an ~80% efficient transformer). That should work in my listening room.




The power supply is straight forward, but you'll notice I'm using silicon rectification. I will snub the secondary (not yet shown on the schematic) to reduce noise a bit more. Initially I wanted to use a 5U4 or 5R4, but the voltage sags too much at full power and drops the output power to something like 7 or 8 watts (both channels driven), so that won't work. Getting a larger power transformer to overcome that voltage sag puts the tube's operating point a bit out of class A. So 450V to 470V high voltage rail is about right for a 5K primary, and a 200 mA 350-0-350 transformer is a very common off the shelf item, so it shouldn't be difficult to obtain. Also, using silicon rectification I can make the first cap larger and get the ripple down so the amp should be dead quiet (but we will see...).



The bias supply is conventional. Nothing extraordinary here. Just need some adjustment range to dial in the tubes to the requisite bias level.

So that's what I've got. This is my first attempt at a single ended design, with no bench prototyping done yet. I welcome any constructive feedback anyone may have on my design.

 

[thorpej]

Looks good to me! That's the nice thing about SE amps -- there's not much to them My first SE amp I did with adjustable cathode bias, but I was only using 6L6Gs. That amp makes about 8W/ch at full power. I sold it to a friend who uses in his home office with speakers that are only 87dB @ 1W/1m, and he's happy with it. I did use a 12AX7 DC-coupled to a 6SN7 cathode follower to drive the output stage, although I probably didn't really need to use the cathode follower because 6L6s are easy to drive when cathode-biased.

 

[gadget73]

Only thing I might suggest is changing the bias adjust method so that if the pot dies the tube goes to full negative grid voltage. Basically two resistors and pot as a voltage divider off the supply. Bias out is the junction of the upper resistor and the pot, and the pot is set up so it increases current through the voltage divider to drop voltage. If the pot dies it just stops dividing.

 

[primosounds]

Looks like it should work. You forgot to add a 0.5 to 1M grid to ground resistor of the 12at7. The power supply of a se amp is the most important part of the equation. I almost always make a clclc or lclc. Since you are also seeking to make the most power from your circuit you should also make it pentode mode. In my experience pentode mode sounds better than UL. But with an OPT that has an UL tap i usually do it that way initially just to make sure that the amp is in working order. The best sounding pentode, supposedly, is using a voltage regulator for the G2 supply.
BTW, Keggar's kt88 design used a pentode driver tube with plate to plate fb. This design is very similar to the much copied (and imo unjustly maligned) RH type amps.

 

[kward]

Thanks for the feedback guys. For biasing I plan on using those sealed 10 turn precision pots if I can find one in the 5K size. Otherwise I'll use a conventional pot and definitely change it so wiper disconnect won't red plate the tube.

Yeah I drew it quickly and forgot the grid leak on the 12AT7.

Pentode outputs seem more complicated with that pesky regulated screen supply that'd be needed. Concerning output power, not sure pentode mode will allow more output power than UL, since UL allows as low a voltage swing as pentode mode for the primary impedance I'm using (at least for the 6550), but now that you mention it, I should go back and validate that. Does anybody know if Pentode outputs would lower 3rd harmonic distortion over UL mode, everything else being equal?

I could use less impedance on the primary to get more current swing and thus more power output, but I expect distortion will be higher.

Thanks again for the responses.

 

[cnpope]

I was slightly surprised when you mentioned substantial power-supply sag at full power. Is that really a significant issue? In class A in the ideal case, the average current demand on the power supply (i.e. average over audio cycle) is independent of the audio power level. Of course an actual amplifier is not ideal, but I would have thought the extent to which the average current draw on the power supply fails to be constant would be a measure of the non-linearity of the transfer characteristic of the output tube. So it would correspond to distortion. Now, I know a single-ended output stage is certainly going to have a significant amount of distortion at high power, but if the non-linearities are enough to cause a substantial increase in the average current draw on the power supply, wouldn't that suggest there would be very high distortion indeed at those power levels?

Also, and presumably related, you spoke of "putting the operating point a bit out of class A." Do you mean the output tube actually goes into cut-off (no current at all passing) during some lower part of the audio cycle , at high power levels? Wouldn't that correspond to an extremely high distortion (and probably normally would be described as clipping)?

 

[dcgillespie]

Neat project! I've had some experience with this type of design a few years ago for a client, so I will be following your project with interest. It used the same OPT, but used cathode bias and a 6DJ8 input tube. As I recall, I ended up having to drop the B+ to 400 vdc in actual practice to maintain optimum Class A operation with that bias scheme. Mine ended up using 9 db of NFB, which produced flat response (+/- 0 db) from 20 Hz to 20 kHz, being down 1 db at 30 kHz. Output impedance was 0.90Ω. Had about 1.5% THD (at 1 kHz) at 6.5 watts RMS in triode mode, which was the mode of primary interest at the time. Had to use a snubber on the secondary to deal with some of the usual Edcor histrionics. All in all, an interesting project.

Also, be aware of Schade feedback in this type of application. It can be effective to be sure, but few of its advocates realize the extreme position it puts the driver stage in, often causing more problems than it solves. That is, the driver tube does not just see the static value of Rfb in this case, but a dynamic value equal to the static value divided by the active gain of the output stage. Therefore, if you apply it to your design as presented, then beside the 47K following stage load, if Rfb is (say) 120K, and the output stage has a gain of about 7, then the driver stage is now seeing an additional dynamic load of about 17K applied to it. As is, the driver stage appears to have little reserve to it. Adding any effective amount of Schade to it will crush it, exchanging output stage distortion for driver stage distortion. Like I said, it can be very effective when correctly applied, but done casually without checking the ramifications to the driver stage, and the results can be far worse with it, than without. Low impedance driver stages with plenty of drive reserve capability need only apply.

Good luck with your project!

Dave

 

[Squidward]

I'd be interested in knowing what you thought about Kegger's KT-88 design that employed Schade feedback, as most of his designs did. I've heard implementations of it that I liked, I was wondering if you felt it suffered from the shortcomings that concerned you.

 

[Dave451]

Watching and learning.....

 

[kward]

I was slightly surprised when you mentioned substantial power-supply sag at full power. Is that really a significant issue?

Thank you for pointing this out, and you bring up a good point of distinction here. For this amp, ripple under quiescent conditions with silicon rectification is 2 mV (that's according to the power supply simulator). Current draw at the power peaks (both channels driven) will cause a 20V peak voltage difference from the quiescent value. With the extra series resistance of a tube rectifier, this difference increases to 33V peak with a 5AR4, 52V peak with a 5U4, and 59V peak with a 5R4. So that voltage difference, or "voltage sag," or what other name you would apply to it, is what I was trying to minimize and hence my desire to use silicon rectification.

I expect this is why others use choke input filtering as that acts as a more natural current regulator which in turn stabilizes the output voltage as power output increases. But I don't feel like adding the weight of a nice swinging choke (if I could even find one). I even considered putting in an actual voltage regulator, such as a Maida type, but would that kill dynamics? Or otherwise it gets a lot more complicated with heat sinking the pre-drop transistor, or adding a pre-drop tube instead of the transistor, like I did on my bench power supply. This all seems too ugly for me to want to wade into for a simplistic SE amp. So best compromise for me was to use silicon rectification with cap input filtering and "live with" the 20V peak voltage sag. Will this voltage sag result in less output power, since the average voltage output is supposedly unchanged? I don't know; I'll find out when I build it!

Also, be aware of Schade feedback in this type of application. It can be effective to be sure, but few of its advocates realize the extreme position it puts the driver stage in, often causing more problems than it solves.

Hmm. Good food for thought. This perhaps explains when I tried Schade feedback on a previous push pull amp with a 12AX7 frontend, that it dramatically increased distortion. I didn't have time to experiment, but it wasn't looking good on the bench when I tried it.

What do you anticipate for bias voltage? I'm not seeing -47v corresponding to 75mA. for 440v. (But that's just me.

One difference you may be noticing is I'm using the GE 6550-A data sheet with UL curves at Ec2 of 450V. I'm using this data sheet because it's available and has a large readable UL curve graph. I'm not sure if the KT88 has exactly the same curves, I expect they're off a bit from the 6550. But for the 6650, it's basically right on the button at 450V plate, 75 mA quiescent current, and -47V grid.

As I recall, I ended up having to drop the B+ to 400 vdc in actual practice to maintain optimum Class A operation with that bias scheme.

That's one thing I wondered about by the way. At 450V quiescent plate voltage, the plate needs to swing clear out to 900V (!!) for true class A operation with a 5K load. But that's clear out there at nearly zero plate current and -90V grid, and I expect distortion might rise a bit out in those weeds. I might be better off backing down the plate voltage and living with a little less output power, or keeping the plate voltage where I've spec'd it and just declare the amp as capable of less output power at a given lower distortion rating. I'm going to prototype one channel on a piece of plywood or something first, powering it with my bench power supply, before I invest in the power transformer.

===

Thanks again everyone for your input! Fun stuff for me exercising a few new brain cells that have been dormant when working on PP gear.

 

[kward]

Also, and presumably related, you spoke of "putting the operating point a bit out of class A." Do you mean the output tube actually goes into cut-off (no current at all passing) during some lower part of the audio cycle , at high power levels? Wouldn't that correspond to an extremely high distortion (and probably normally would be described as clipping)?

Correct. If you increase the plate voltage while readjusting bias so you don't idle the tube hotter than it can stand, you create the ability for more signal swing on the positive going input signal, but trying to swing that same amount on the negative going input signal will cause the tube to cutoff. Those are conditions that you'd like to have for a push pull amp, though.

 

[dcgillespie]

A Class AB1 push-pull amp that is.........

Dave

 

[primosounds]

Correct. If you increase the plate voltage while readjusting bias so you don't idle the tube hotter than it can stand, you create the ability for more signal swing on the positive going input signal, but trying to swing that same amount on the negative going input signal will cause the tube to cutoff. Those are conditions that you'd like to have for a push pull amp, though.

Kev, i use fixed bias on 2 of my hi power SE amps and in fact i consulted with you way back, when i first started the project about the circuit design. I am using mil spec type AB pots, which are very nice and are almost sealed. I don't anticipate any problems with reliability over the long run. The pots made in Canada are also very good too. I tried to add the fail safe resistor but it made getting the desired negative voltage in much smaller range which required changing resistor values. In the end i felt it was better to have the greater voltage range and left off the safety resistor.
Anyway, regarding RH circuits, Alex Kitic, was very adamant about following the circuit as he designed and did not like it when people deviated. He did a lot of Spice work to find the components values he settled with. At the time he did not understand that in the 'states' people like to improvise or customize and wanted to try different tubes and different values of the circuit which of course led to vigorous discussions!

 

[cnpope]

Correct. If you increase the plate voltage while readjusting bias so you don't idle the tube hotter than it can stand, you create the ability for more signal swing on the positive going input signal, but trying to swing that same amount on the negative going input signal will cause the tube to cutoff. Those are conditions that you'd like to have for a push pull amp, though.

Yes, but in a single-ended amplifier, surely pushing the power to the point where the tube is cutting off at the bottom of the audio cycle means you have pushed to the point of clipping, and appalling distortion?

Coming back to the issue of the power supply sag, isn't it the case that the extent to which this happens, in a single-ended amplifier, is a measure of the extent to which the amplifier is distorting? (If there were no distortion, the upward excursion in current in the positive part of the audio cycle would be exactly compensated by an equal and opposite downward excursion in the negative part of the cycle.) I'm not sure how much distortion is considered acceptable in single-ended amplifiers, but it does surprise me a bit if so much distortion is tolerated that the average current draw from the power supply changes appreciably, thus implying an appreciable sag of the supply voltage. (Of course, the fluctuations during the course of the audio cycle should be evened out by the use of sufficiently much capacitance in the supply.)

 

[dcgillespie]

In an ideal single ended Class A amplifier, the current drawn by the output stage should in fact be utterly constant. Reality has it fluctuate a bit, but even then, with good design, it can get pretty close. But this is always relative to the signal at the plate, where the current should be constant. When a screen grid enters the picture however, fluctuations are more pronounced -- and particularly so with true pentode tubes -- as the screen's current draw is not nearly as linear as that of the plate's, relative to the applied signal.

Dave

 

[cnpope]

In an ideal single ended Class A amplifier, the current drawn by the output stage should in fact be utterly constant. Reality has it fluctuate a bit, but even then, with good design, it can get pretty close. But this is always relative to the signal at the plate, where the current should be constant. When a screen grid enters the picture however, fluctuations are more pronounced -- and particularly so with true pentode tubes -- as the screen's current draw is not nearly as linear as that of the plate's, relative to the applied signal.

Dave

Thanks. That is very helpful, and resolves my puzzle!

 

[dowto1000]

Hello, here is strictly "my " take, YMMV, and you asked for the ideas of others :

A KT-88 ( Gold Lion or Genelex ) is a superbly FINE tube choice for finals in a SE amp, one of ONLY three tubes I'd personally build with, in SE amps.

***** A number of years ago, the principal at QuickSilver , Roger Sanders, brought a two stage ( 1/2 12AX7 - KT-88 ) SE amp to RMAF, and it was fabulously good sounding, fabulous. Research THAT design !! It was in the Azzolina Audio room that they A-Bed it against a bunch of other amps. It used a REL Teflon coupling cap, small value, about 1 inch long and 5/8ths inch diameter, and it had a TINY iron core choke on the Driver tube's plate as I recall. Roger KNOWS what he is doing !! ********

Speakers, Azzolina's latest, are here...interesting ..http://azzolinaaudio.com/

I would avoid complex circuits - three stage amps like the plague. No cathode followers.

The higher the mu driver in a two stage amp, the LIVELIER the sound and more mentally-engaging the amp. I love a mu of 100 driver - fun to hear. How about a 12BZ7 driver at 190 Ea and 2.2 mA?? Don't worry about that tube's odd curves, or distortion at maximum swing, 98% of the time, it is a small swing. Low Rp for mu of 100. Great sonically.

I would not design for maximum power, but rather, maximum sound quality.

I am very sure you have grossly over-filtered your B+ to the finals tube.

The amounts of C you chose, 220 uF and 550 uF is far too high, it will rob the amp of being able to do its best in dynamic contrasting. On VOTTs at my home ( about 100dB ) I can run up to 800 mVAC peak to peak, ( PSUD ) and not have a huge hum problem, whereas you are filtering down to just a few mVAC.

When you use large C to the Finals, on " recharge time " of any large transient, the much-=lower Z "C" hogs and gets the recharge, and not the tube. Equals LESS instantaneous peak dynamic contrasting, and the SE amp becomes artificial-sounding ...like most others are, ( ask Bill Wojo ) less-real sounding.

This means, in a SE amp, one uses TWO L/C sections in series. No large high DCR chokes and large Cs in any one position to filter the finals tube. IF this is a stereo amp, for a minimum in quality, use a shared L1/C1, and a separate L2/C2 for each channel to the finals. Especially with a current-heavy KT-88.

There is no other way I am aware of, to pull off the KT-88 stereo amp design. It becomes L1/C1/L2/C2 ... with a "Y" and a separate L2/C2 for each channel.

A KT-88, my friend, because of its high current, in stereo, is HARD to design for thoughtfully. I won't attempt it.

Sonically, I like the sound of directly heated tube rectifiers.

Instead of running the KT88 at 80% of maximum, reduce current ( not VDC P-K ) until total KT88 plate dissipation approaches a Golden Ratio percentage, about 62 %. This will ensure LONG reliable finals tube life, cool operation, and it avoids a thermally stressed "sound" I hear, when tube plates are run "hot".

If you were to direct couple the driver to the finals tube, you employ a larger value Rk on the finals ( 3K to 5K ) and so. it takes a smaller Rk bypass cap, which CAN ( and should !!) be of high quality film types. Use a WIMA DC LINK as the main Rk bypass cap, 4 lead versions, not 2 lead.

A fixed bias supply has to be PRISTINE in its design, parts choices, and operation. It's parts signature is superimposed right on the grid. I see no need for it in your amp, the added extra complexity, and it is failure prone VS self bias. A well designed SE amp should have NO user adjustments, pots, hum pots, etc.

It would be easier to keep the speakers you have, purchase a pair of 98 dB Seismic 15 inch coaxials, and build a proper sized SE amp, 3 Watts or maybe less. Look up that 15 incher !!

Talk ASAP to the gentleman at QuickSilver about HIS SE KT-88, he once took to RMAF and the Azzolina room. .

http://quicksilveraudio.com/

Have fun, I love your attitude and openess !!

Dowto1000

 

[Tubeglowpio]

Thanks for the feedback guys. For biasing I plan on using those sealed 10 turn precision pots if I can find one in the 5K size. Otherwise I'll use a conventional pot and definitely change it so wiper disconnect won't red plate the tube.

Yeah I drew it quickly and forgot the grid leak on the 12AT7.

Pentode outputs seem more complicated with that pesky regulated screen supply that'd be needed. Concerning output power, not sure pentode mode will allow more output power than UL, since UL allows as low a voltage swing as pentode mode for the primary impedance I'm using (at least for the 6550), but now that you mention it, I should go back and validate that. Does anybody know if Pentode outputs would lower 3rd harmonic distortion over UL mode, everything else being equal?

I could use less impedance on the primary to get more current swing and thus more power output, but I expect distortion will be higher.

Thanks again for the responses.

Can't wait to hear what you think, be careful not to use those blue 10 turn pots off of eBay. Before I wired them up I checked them out and they liked to jump all over the place and fully closed and opened at certain spots.

 

[Tubeglowpio]

Neat project! I've had some experience with this type of design a few years ago for a client, so I will be following your project with interest. It used the same OPT, but used cathode bias and a 6DJ8 input tube. As I recall, I ended up having to drop the B+ to 400 vdc in actual practice to maintain optimum Class A operation with that bias scheme. Mine ended up using 9 db of NFB, which produced flat response (+/- 0 db) from 20 Hz to 20 kHz, being down 1 db at 30 kHz. Output impedance was 0.90Ω. Had about 1.5% THD (at 1 kHz) at 6.5 watts RMS in triode mode, which was the mode of primary interest at the time. Had to use a snubber on the secondary to deal with some of the usual Edcor histrionics. All in all, an interesting project.

Also, be aware of Schade feedback in this type of application. It can be effective to be sure, but few of its advocates realize the extreme position it puts the driver stage in, often causing more problems than it solves. That is, the driver tube does not just see the static value of Rfb in this case, but a dynamic value equal to the static value divided by the active gain of the output stage. Therefore, if you apply it to your design as presented, then beside the 47K following stage load, if Rfb is (say) 120K, and the output stage has a gain of about 7, then the driver stage is now seeing an additional dynamic load of about 17K applied to it. As is, the driver stage appears to have little reserve to it. Adding any effective amount of Schade to it will crush it, exchanging output stage distortion for driver stage distortion. Like I said, it can be very effective when correctly applied, but done casually without checking the ramifications to the driver stage, and the results can be far worse with it, than without. Low impedance driver stages with plenty of drive reserve capability need only apply.

Good luck with your project!

Dave

I've been thinking about building the SE kt88 from diyaudioprojects, sounds similar so sounds like a promising project. I would also be using a 6dj8 in that circuit but would change the power supply a bit and maybe optimize it for a kt120. I need at least one se amp with some power.

 

[Tubeglowpio]

Hi,


You state:

" I need at least one se amp with some power.

"


What you need is amps that play dynamically. This means NO chokes 50 Ohms or more, to the Finals, and ideally, 20 Ohms or less. It also means attention paid to lay out, wiring types, wire lengths, and wire AWGs, so as to maximize transfer efficiency in the SE amp, below deck. The good battle.

As you go to higher current amps, you run into SE design problems, assuming you are seeking best sonics and performance.

Dowto1000[/QUOTE]
I don't mind my serious listening amps to be low powered because it is easier to find better parts but I would like to try a higher powered se amp that could rock my 96 db speakers.