• Disclaimer: This Vendor's Forum is a paid-for commercial area. Unlike the rest of diyAudio, the Vendor has complete control of what may or may not be posted in this forum. If you wish to discuss technical matters outside the bounds of what is permitted by the Vendor, please use the non-commercial areas of diyAudio to do so.

UNSET is coming?

UNSET is coming......

I have mentioned these words a few times in passing, but the truth is that UNSET only existed in LTspice at the time. The simulation showed unreal performance, but that is often the case with vacuum tube simulations, especially those where the tube is not connected up in a conventional manner.

This was the case for about a year. I have too much on my plate to effectively handle right now, so the simulation, along with several others sat in my PC undisturbed for a long time. Late last year I put together a test board for a push pull amp that used UNSET technology in the output stages only. It did work rather well, and I planned to pursue a follow up design, but the need to redesign the TSE board popped up, which is not yet finished. The boards are done, but all the documentation is still a mess.

About 3 weeks ago my wife started exhibiting heart attack / stroke symptoms, but refused to go to a doctor until bad stuff happened. There were good days and bad days for two weeks, which required my full time attention. During the evening of one of the good days we were both at this computer ordering some stuff for a baby shower when she said that she didn't feel well and wanted to go upstairs. Before we got to the stairwell, she passed out and dropped to the floor. I managed to get her upstairs, out to the car, and to the ER. They stabilized her and eventually figured out what the cause was, but this was the beginning of 4 days in the hospital and some full time watchfulness from me once she was home.

During the 4 days in the hospital she was sleeping most of the time so I fetched my laptop and did what I often did with a few days of idle time.....took one of my best simulations and made it into a PC board.

UNSET the simulation became UNSET the test board. it was a hurried up layout, many parts didn't fit right and a few were just left out by mistake. Over the past few days I populated the board with the exact parts from the simulation except where I didn't have the right part, I stuck in whatever I had. What were the chances that this would work? Would it blow up in my face? Am I going to stick expensive tubes into this thing? No way........

So, briefly, what is UNSET?

A SET is a Single Ended Triode amplifier. The SSE, TSE, and new TSE-II amps are examples of this, but the SSE often uses pentode tubes wired as triodes. This works, the pentode takes on triode like qualities with the associated triode disadvantages, most notably the inability to pull its plate down near the cathode voltage thus limiting the available power output. Another issue that needs to be overcome is the screen grid voltage limitations of most TV sweep tubes. Wire them as a triode, and most will eventually blow up when left alone idling which is worse case for a class A amp (maximum dissipation). The tube doesn't know if it's pentode, UL, or triode connected when it's idling, it just gets a constant voltage on it's elements, and if the screen grid is significantly above it's maximum spec, there is a risk that it will blow up (some will, some won't, but most will if left long enough).

What if there was a better way?

I have been searching for this better way for nearly 10 years. I experimented with screen drive, dual drive, and even built my GUT (Grand Unified Theory) board which had the ability to drive every electrode independently in either phase, except the plate and heater. Several toasted (and two exploded) tubes and nearly 10 years later, I arrived at a new topology that I can't find anywhere it recorded vacuum tube history.....yes, there are several close similarities, but this is truly unique.

I called this topology the Composite Electron Device for lack of a better name, since it is a composite of a vacuum tube pentode, a mosfet, and a hand full of discrete parts to create triode like curves. The screen grid specs are not violated, so that nearly any pentode will work with appropriate component values. Note that this statement is still based primarily on simulations, with some limited actual tube testing.

I will not reveal the full schematic just yet, but it is coming.........

The UNSET is an amplifier design that uses CED topology for its driver and output stages. A small CED is built with a high Gm pentode for the input / driver stage. I used a 6EJ7 / EF184 in the simulation, but stuck a $1 tube in the test board.

The output stage uses an identical circuit built with bigger parts. I laid out the board such that several popular octal tubes could be used, but messed this section up pretty bad. I wired it up to use an unloved and unwanted octal sweep tube that I have LOTS of.....A popular tube seller sold me over 100 of them for under 50 cents each!

After finding my mistake with a backwards zener diode, I was really surprised by what happened next.........
 
Last edited:
I connected the board up to an external bench power supply, my old Fluke 407D. I installed one input / driver tube and attempted to test the input stage by itself, but voltage readings everywhere were all wrong. Nothing came anywhere close to the simulation, and the output was highly distorted. Some poking around with a voltmeter revealed a mistake I often make.....I got a diode in backwards in both channels.

After fixing the backwards diodes the input / driver stage fired right up and revealed performance close to the simulated circuit. The DC voltages were a bit off, but I didn't even have the right tube in the board, so how could I complain.

I cranked up the input until the driver hit clipping and backed up a wee bit. I was getting 56 volts RMS! This is more than enough to drive the grids right out of just about anything. The THD at a sane 20 V RMS was 0.2%, no complaints here. Frequency response? What's that? The driver is flat from 5 Hz (generator limit) to somewhere beyond 100 KHz where the 8903A stops reading.

I stuck in a pair of 25DN6 tubes and wired their heaters to a 24 volt SMPS, turned up to 25 volts. The 25DN6 is an old sweep tube dating back to the mid 50's. They can still be found in large quantities for about $3. They are NOT all created equal. The old ones from the 50's are built to their 15 watt plate dissipation ratings. Some of the newer vintage tubes have BIG plates inside! The GE's have a plate that is identical to what's inside their 24 and 28 watt tubes, so that's what I stuck in the amp.

I set the Fluke on 450 volts, and turned the bias up to 100 mA per tube.....yeah that's about 40 watts (OPT and other losses). I used a pair of 3000 ohm Transcendar OPT's that were still on the bench from my TSE-II board testing. The amp came to life immediately and surprised me with performance.

The THD at 5 watts was under 1% except at the frequency extremes where the OPT limited the performance.

The -3dB point at the low end was 8 Hz where the THD was 26% due to OPT saturation. At 20 Hz it was 0.54 dB down at 4.03% THD.

The upper -3dB point was arounf 50 KHz where the THD was 1.58%. At 20 KHz the amp was 0.62 dB down at 0.56% THD. NOT bad......so crank it up!

Thd was 0.197% at 100 mW, rising to 0.235% at 1 watt. 5 watts brings 0.662%, 10 watts 1.61%, with 2.48% at 15 watts and 4.04% at 20 watts clipping sets in at 20.8 watts where the THD hits 5%.

I then hooked up some speakers and cranked on it for several hours. It's one of the best sounding amps I have heard.....I really want to do a side by side with a TSE-II.....

I did discover that the amp will exhibit a foldover situation when both channels are driven well into clipping. This was traced to my conservative settings on the current limiter in the screen regulator which feeds all 4 tubes. It would be easy to just turn up the current, but maybe two separate regulators is a better solution......time will tell.
 

Attachments

  • P2920681_x.jpg
    P2920681_x.jpg
    834.1 KB · Views: 3,000
I hope Sherry is feeling better now, scary stuff ...

Really looking forward to trying this - sounds like maximum output is roughly about 85-90% the plate dissipation rating of the tube. Quite an improvement on the usual 20 - 30% ...

Will we need a screen voltage? Any advance information on power supply voltages, or other special stuff we will need to be prepared for this board, or is that too much of a reveal at this point in the development?
 
Will we need a screen voltage?

The current test board has a screen regulator (zener stabilizer for those who like to argue the details) to derive the screen supply from the main B+. All other voltages are derived on board.

The typical board as I see it will require a transformer like the SSE, a HV winding with a CT, a 5V winding if a tube rectifier is used, and a heater winding. Some big fat sweep tubes need as much as 2.65 amps each (6LW6) and the two driver tubes need 300 mA each. Since there are LOTS of odd voltage TV sweep tubes out there the heater connection for the output tubes are brought out to a separate connector so that a separate transformer or SMPS can be used for them. Most of the usual driver tubes are available in 6.3 volt and lower heater voltages. The two heater connectors can be wired in parallel if 6.3 volt output tubes are used.

Any advance information on power supply voltages, or other special stuff we will need

After testing some amps with the UD board which require several power supplies, I gave some serious thought into simplification. The "dream" goal was to get TSE sound quality from a simplified power source and cheap tubes........300B = $100+, 25DN6 = $3, 36LW6 about $15. I really need to stuff one in there and see what I can get out of it on say 600 volts, if that much doesn't set an OPT on fire!

One of the next steps will be to hook up an Antek toroid.

sounds like maximum output is roughly about 85-90% the plate dissipation rating

Not really. Yes, the 25DN6 has a 15 watt plate rating. The data sheet was written in 1955. Tubes produced in the late 60's and 70's all used a standardized plate structure. The GE 25DN6's that are in there now have the same plate found in their 24 and 28 watt sweep tubes, which is larger than the plate found in a 6550. They just show a hint of red in a dark room at about 45 watts. Most of the RCA 25DN6's that I have have the same plate structure as their 6DQ5. I haven't tried them yet to see their true capabilities.

I would guess that this amp gets close to the theoretical maximum 50% efficiency, so I'm putting 40 watts into the tube to get 20 watts out.....that's when it is cranking out a pure sine wave at just below clipping. Turn the volume down to zero and the tube eats ALL of those 40 watts.....that's the true limiting factor in any SE amp that runs true class A.

I didn't mention this before, but as currently built there is no global feedback applied anywhere in the amp. Each stage has the local feedback loops needed to make triode curves from pentodes. Each stages operates completely independently from the other. I plan to pull the driver tubes, connect a push pull OPT to the output stages, and drive them with a UD board, but that's another project.

A global feedback resistor can be applied from either side of the OPT if desired. This shows a serious further reduction in THD in the simulation, but has not yet been tested in the amp with music running through it. It might help squeeze more sound through some budget OPT's due to its lowering of the output impedance. I still have a pair of Hammond 1628SEA's that might eat 600 volts and pass 30 to 40 watts............
 
I hope Sherry is feeling better now, scary stuff ...

She is doing much better now, thanks. This mess started with an unusual spike in her resting blood pressure, from normal to quite high, which drew the usual response from her cardiologist......up her meds. A couple weeks in Florida coupled with double doses of her blood pressure meds did not reduce the BP, it kept climbing. So upon return they added a diuretic, which along with some dehydration from the Florida heat caused her electrolytes to drop to "near coma" levels. 4 days of IV drips brought them and her back to a functioning, but not completely normal condition. The BP problem remains however.

On top of that our youngest grandkid, age 5, shattered his wrist yesterday. The fun continues.

Back to the UNSET.....I listened to it for several hours last night, still quite amazed by it's stunning realism, but in true Tubelab style, I had to mess with it.......now it's dead.

It started when I swapped the Transcendar OPT's for the big Hammond SEA1628's which have never sounded quite right when fed by the usual suspects, a 300B, or triode wired KT88, so would a triode emulating sweep tube do any better? The transcendars are 3000 ohms, the Hammonds are 5000 ohms, but could be wired for 2500. I chose 5000 for the first test.

I had seen the screen supply collapsing due to a built in current limit when driving it hard with music, but it seemed worse when I drove both channels into clipping with a 1KHz sine wave, so I did the usual Tubelab quick fix. I put a clip lead jumper across the sense resistor.

Power output went way up, but before I could record any numbers there was a bang, and some mosfet parts scattered on the workbench. The amp however kept on making power, so I believe that the mosfet is probably the only casualty, but my sense resistor had unsoldered itself, so the fet obviously needs a bigger heat sink. I'll get back to if after it cools off.
 
Mischief grand pa style

Not intentionally. Apparently there were random clothes scattered on the floor in the room he shares with the 7 year old. He tripped on something and fell, whacking his wrist on part of the bed frame or a night stand on the way down. The same kid broke his elbow last year by crashing down the stairs. The other three kids have had no issues, so he must be clumsy, fragile or both.

Power has been restored to the UNSET board. The mosfet was scattered, but it was the only casualty. I had soldered a resistor in parallel with the sense resistor in my first attempt to boost power, which I subsequently jumpered. The resistor became unsoldered and the jumper which was connected to B+ left its mark on some small parts, but none of them were blown. A bigger mosfet and a 5 watt sense resistor have allowed for continuous operation well into clipping. Some quick knob turning seems to indicate that the big Hammond likes this circuit and it's HF response extends to beyond 30KHz.

I have also discovered that the distortion figured on my 8903A are meaningless random numbers when used below 20 Hz. The amp was measuring 0.5something percent THD at 20.5 Hz but transitioned to 40 something percent when the knob was bumped to 19.5 Hz. The scope display continued to show a perfect sine wave. More testing tonight maybe.
 
I got one more experiment in today.

The date on these Hammond OPT's is 5-11-06. Yes I have owned them for 13 years and never found a suitable way to get HiFi quality audio through them. They can be seen here with one of the first SSE prototypes powering up some Lowther based horns. They sounded terrible here since the 106dB efficient Lowthers were loud with half a watt. The final "first SSE" wound up using a pair of Transcendars like those I just pulled OUT of this amp. They sounded quite nice with the big horns.

The power at clipping had dropped to about 15 watts when I traded the 3000 ohm Transcendar OPT's for the big 5000 ohm Hammonds. How do I get more power......turn the knob to the right!

I cranked up the power supply to 520 volts. That's as far as I want to push the 500 volt filter caps that are in the board. This brought 22 watts at the edge of clipping (5% THD @ 1 KHz). THD at 20 watts was around 2%. I backed the power down to 10 watts and turned the frequency dial on the 204C. At 1 KHz the THD is 0.68%. At 20 Hz the output drops 0.40 dB and the THD rises to 0.88%. Going much below 20 Hz starts making my all original Fluke 407D angry. It's current meter is hovering around 260 mA.

Spinning the dial in the other direction reveals a drop of 1.01 dB at 20 KHz. This is where the famous Hammond notch of anywhere from 5 to 15 dB resided in the past. This board has "tamed the beast." Turning the knob past this point sends the current meter on the Fluke beyond the end of it's scale, so I didn't test there. The OPT's get very capacitive beyond that notch driving the tube current up, and into the red zone.

Running the output stages at 520 volts and 105 mA brings on some mild red glow, so I must point out another feature in the UNSET board, albeit one that I botched pretty bad in the layout.....the "Any Octal Matrix." The board can be configured for almost any octal output tube from the little 6V6 and 6AU5's to the fat boy 6LW6......so you know where I'm going next. That's right, bigger tubes and higher voltage caps.
 

Attachments

  • SimpleSE_640.jpg
    SimpleSE_640.jpg
    272.8 KB · Views: 1,525
  • The_Amps_2_640.jpg
    The_Amps_2_640.jpg
    318.4 KB · Views: 1,508
  • Ampsa.jpg
    Ampsa.jpg
    245 KB · Views: 1,552
  • P2920684_x.jpg
    P2920684_x.jpg
    891.6 KB · Views: 1,502
... I cranked up the power supply to 520 volts. That's as far as I want to push the 500 volt filter caps that are in the board. This brought 22 watts at the edge of clipping (5% THD @ 1 KHz). THD at 20 watts was around 2%. ...

How much power can be obtained from a big DHT in SE, like SV 572, 211 or 845?

This reads like it is putting out as much power, at 1/4 to 1/3 the plate volts, and less distortion, as the big DHT's, and you haven't even gotten to the big sweeps yet?

I guess the 6146 family would work here like any other pentode sweep tube.

Dumb question: will this topology lend itself to being wrapped around real DHT's to linearize / enhance them?

edit: will there be any provisions for compactrons, or will we need to make an adapter?
 
Last edited:
Hey Win - you just reminded me of a massive stash of NOS 8552s (a 6146-oid) I have somewhere down in the basement. I "inherited" them when a volunteer fireman friend was cleaning out his fire station's equipment room. They're not good for much else, so maybe they'd work here.

Woo-hoo! :)
 
How much power can be obtained from a big DHT in SE, like SV 572, 211 or 845?

Never met an SV572, but I do have an amp that can run a 211 or 845. It uses a first gen TSE board with a 45 tube that feeds a big PowerDrive board that has no problem driving anything into A2. It was used to extract 200 watts of SE power from an 833A on 1500 volts. With 211's or 845's the power at the edge of clipping is just over 30 watts. The 211 has a higher Mu and is easier to drive, but gives a higher output impedance which doesn't work well with some speakers. I prefer the 845 due to its better bass with my speakers.

Coincidentally the power rating on the big Hammond OPT's is 30 watts, so that's the target power level, and I believe that I can get there with a big fat sweep tube, or maybe two smaller ones.

I planned to test that theory today, but Sherri vetoed those plans. She will be away for a couple of hours tonight, so I plan to audition those OPT's with the existing board at a high volume level:)

will there be any provisions for compactrons, or will we need to make an adapter?

That brings me to another decision point, and one of the reasons for not just ripping out the power supply caps and turning the B+ voltage up beyond 600 volts. The CED circuit design uses a mosfet and a pentode combined with some other parts to create the "triode." The mosfet will need to dissipate 5 to 20% of the power seen by the pentode. This means that a 40 watt sweep tube that's turned up into the 50+ watt region may need to burn 10 watts or so in the mosfet. That's not wasted power, it does get turned into audio with efficiency about the same as the pentode.

Yesterday's excursion to the 520 volt zone resulted in 44 watts being dissipated in the tube at idle, and 9 watts being dissipated in the mosfet. This drops as the amp is driven and a lot of that energy gets turned into audio power, 22 watts at full crank. This means different size heat sinks for different power levels. Yesterday the tubes had a dull red glow at idle, but the heat sinks were too hot to touch. Going beyond 500 volts means bigger heat sinks as well as bigger caps. There isn't room for bigger heat sinks on the board, but provisions for top or board mounting are included.

My initial thoughts, and this test board, put everything on a board that's about the same size as a TSE-II. The idea was to design two, maybe 3 boards. One for octals, one for 12 pin compactrons, and maybe one for 9 pin Novars and Magnovals. The heat sinks that are used here are good for power levels up to 15 watts or so, and higher power levels could use the chassis for heat sinking.

Maybe its a better idea to go with the modular approach with separate smaller boards that can be wired together to build amplifiers of any size. The mosfet would be on the board with it's associated tube, but could be mounted on on top or bottom, possibly in one of two locations.

Any thoughts / preferences.

will this topology lend itself to being wrapped around real DHT's to linearize / enhance them?

I'm not sure exactly how this would work since the screen grid is a driven element, and part of the triodization process. I'm sure that the concept could be applied to a true triode, but I haven't gone down that road yet. There were far too many pot holes and speed traps on this one.

you just reminded me of a massive stash of NOS 8552s

I had about 50 6159's but gave them all to a ham that wanted to build a big linear amp. I probably still have a few 6146's in a box around here somewhere. I have dozens of tubes I want to stick in this thing ranging from 6V6's to big fat sweep tubes, but it will take a while to burn through and document them all. It appears that the triodization process requires some different parts resulting in different drive levels for each grid. This may need optimization on a tube number by number basis. I have had good luch with small and medium sized sweep tubes so far. Just sticking a KT88 in place of a sweep tube didn't work well, but is has a much less sensitive screen grid.....time and testing will tell.
 
I cranked up the power supply to 550 volts (max output for my old Fluke 407D), pointed a small desk fan at the heat sinks, turned up the tube current to 120 mA (max spec for the Hammond 1628SEA).

Power output at edge of clipping (5% THD) rose to 25 watts.

Power consumption breaks down as follows:

The B+ voltage is 550 volts. The current meter on the power supply reads 270 mA for a total DC input of 148.5 watts. 2.6 watts are lost in the OPT's DCR. The mosfet eats 10.6 watts, and the tube burns 52.8 (yes it's rated for 15 watts. I promise I didn't melt them....... This is a total power input of 183.5 watts, but some power will be wasted turning line voltage into 550 volts DC, and some will be lost in the SMPS for the 26 volt heaters if used. I'm assuming that total power draw will be in the 200 to 250 watt range. The amp is class A, so the power consumption doesn't change much from idle to max power.

Scale that up to 30 watts output (maybe 600 volts B+). The heater consumption doesn't change, neither will the B+ current due to the OPT max rating, so the B+ power consumption goes to 600 volts * 270 mA = 162 watts, and the total goes to 197 watts. This looks like 225 to 275 watts drawn from the line.

In contrast that to my 845 SE amp. It ran 80 to 110 mA per output tube at about 1050 volts DC. We will call it 100 watts per output tube. This was derived from a backwards wired 480 volt industrial control transformer feeding a voltage doubler made with 5AR4's (10 watts per heater). Total power from the wall was over 300 watts but I don't remember the exact number. Their heaters draw 3.25 amps from 10 volts EACH and they need DC to be quiet. That's another 80 or so watts.

The driver was a complete 45 based TSE with a 450 volt power supply to feed the CCS load on the 45 (no OPT, just a powerdrive board). I don't remember the exact numbers but the complete amp drew well over 500 watts from the wall, and really heated up my room, so it didn't get much use in Florida.

So after turning up the power, I connected up some speakers and let it play for about 3 hours......ME LIKE!

After a couple hours I began to smell something toasty I was thinking the 500 volt cap eating 550 volts, but it turned out to be my old Fluke 407D. It didn't like running at max power for 3 hours, so I shut off the fun. I will be gone most of tomorrow and Sunday.


I sat there staring at the board and sketching out stupid ideas on scratch paper while I cranked out the tunes from the amp tonight.....here are my latest ideas, but I may change my mind by tomorrow. I'm really thing about a short path to building myself one of these with some chassis and stuff that I already have. I have the Hammond OPT's and some big Antek power toroids.....

I kind of like the idea of a hole in the board, and pads to skywire a builders choice socket to the board.

That still leaves the issue of multiple different heat sink sizes. I'm now thinking of one board similar to this one with all 5 tube sockets and heat sinks like what's on the board now. It could work at power levels up to 10 watts and use any of the common older generation sweep tubes. Sort of like a cross between a TSE and a SSE.

Then for us experimenters and power hungry amp builders a second smaller board, ditch the rectifier tube since a 5AR4 will not support the power levels needed for a 30 WPC SE amp, so silicon (on board) or damper tubes (off board) must be used. There would not be any output tubes on the board either, just terminals for the 5 electrodes being used. I got visions of those 4D32's wired into this thing. The heater wiring could go to the board, or directly to the transformer or SMPS, with one terminal for the variable heater DC float voltage. The mosfets would be mounted on (or to) the board's rear edge to allow multiple mounting schemes.

I have a little push pull version where I did stuff everything on one board. It works really good, but gets too hot and was a nightmare to build. The heat sinks are on the bottom. It used a conventional pentode LTP for the PI / driver since the CED does not lend itself to being wired as a PI. The output stage is CED with a tiny 6GF5 sweep tube with a 9 watt plate rating. 50 WPC flows easily without and redness from an Antek power toroid, and 70 WPC can be had with the Fluke feeding a regulated 500 volts.
 

Attachments

  • P1850379_x.jpg
    P1850379_x.jpg
    548.8 KB · Views: 1,091
  • P1850383_x.jpg
    P1850383_x.jpg
    569.5 KB · Views: 548
  • P1850371_x.jpg
    P1850371_x.jpg
    451.1 KB · Views: 578