Since the UNSET is still a prototype / beta build, there are no real specs, especially since every one is built differently. The THD in any SE amp without feedback will never be in the range of a P-P solid state amp with 40+ dB of GNFB.
That said, 1.2% at 1 W seems a bit high as I have been able to get well below 1% at 1 W in most of my experimental builds. I do routinely get readings from 0.4% to 0.7% on my UNSET experiments at 1W and 1 KHz. It took a lot of messing with my bench setup to get there since things like ground loops and even LED bench lighting affect the readings. Still the Walmart LED shop lights over the bench raise the THD about 0.1% when they are on.
9 watts at %5 is also a bit low since I was getting 12 to 15 watts at 3 to 5% with a 6DQ5. What is your B+ voltage and OPT impedance?
The 6KT6 is a typical IF amplifier from a TV set. Unfortunately, there are a lot of different "things" out there with the type number 6KT6 screened on them. I have several hundred "6KT6" tubes marked GT. BRITTAIN. They came in sealed bulk packed boxes marked in oriental characters, so I doubt they are really British tubes, and I'm not sure that they are 6KT6's either. I have lots of them so I use them for testing. I have about 5 other different tubes marked 6KT6, and again they are all different. All work in the UNSET, and in some of my other experiments including a vacuum tube music synthesizer project. I have tried every tube that I can find that has a compatible pinout in the UNSET board, All work, and as expected, some work better than others. Each has a different "sweet spot" for gain and THD performance. All were simply plugged into the board with the pot adjusted for best THD. I have not yet tried to optimize the component values individually for each tube. So far, the higher Gm tubes seem to work better.
The bench rebuild project seems to be dragging on. Sherri wants all the Christmas decorations up by Thanksgiving, so that has been taking most of my time.....come on, who needs 5 Christmas trees?
I have not yet put power to the bench, so nothing has been tested. I am running a dedicated supply line from the breaker panel to the bench, and moving the WiFi router off the bench since I have seen it do funny things to my measurements. It's currently about two feet from the test amp setup, and about a foot from the audio analyzer. The LED shop lights affect the THD readings, so they are being moved further away. I have had to unplug the router and turn off the lights in the past to get steady readings.
The previous setup used an HP8903A audio analyzer for THD measurement. I have completely torn that setup down, and I am now adding a PC with a sound card and REW to the bench and rearranging everything including lighting and power distribution. I'm sure that there will be some serious ghost hunting to chase all the residual THD adders out of it, since the PC itself is often a big contributor. The old PC I am using now has an internal 24 bit 192 KHz card inside it. It's the same one that I used in my lab in Florida for 10 years. It took some tweaking to get all the gremlins out of that setup, and I still had to live with two spikes on the FFT plot from the PC's CRT monitor before switching to LCD. If there are insurmountable problems, I can always rip the Focusrite Scarlett Solo off of this PC and use it there, but that means messing with two computers. I expect a similar setup shakedown on this bench, but it will offer much more useful info and handle more simultaneous experiments once it is done.
I have about 15 8068 tubes. All vary from used, to abused, as they were pulled from dead power supplies. Some say that these are just a 6BG6 with the plate from a 6CD6 in it. A few of mine do look like that, but most have a larger plate with more room inside it than what's seen in any 6CD6. The 6L6 / 6BG6 cathode limits the average current to about 100 mA, and the peaks to maybe 300 - 400 mA, so these are no good for a big P-P amp, but the low heater current draw may make them a good candidate for a small UNSET that generates minimal heat. Testing is required. As Parafeed813 says, to be continued......
Crusty old used 8068's don't fetch much $$$$, especially those that were pulled from scrapped Kepco's, so mine will be tested in an audio amp.
I have zero EL36's but I do have a box full of E130L's somewhere. They were also pulled from "power supplies" by a metal scrapper as were the 8068's. I saw, and purchased a few of the Kepco's, but I have no clue as to what the E130L's were actually in. Most are "Made in Hungary." I did try some in a P-P test amp with good results, so I kept them. Once found, they will meet the UNSET board.
Another one powered up today.
6JC6 into 6CD6 - chosen because I'd seen them pictured in Tubelab's photos and figured following in the master's footsteps was the surest way to success without glitches.
Had a couple of minutes figuring out why the outputs weren't conducting until I realized the current sense 1Ω resistor pads for short and long bodies weren't joined. Then up and running without any trouble at all .
Input tube voltages are currently at 123Vdc and 6CD6 outputs are both fairly steady at around 55mA , wandering a little by about +/- 5mA.
B+ is roughly 480Vdc. Interesting that the output currents seem to be tracking together, maybe as line voltages fluctuate.
One comment : It only took a couple of minutes working on the board to get a feel for the years of experience that went into it. Even pad/hole size relationships made the placing and soldering . . . . I dunno, just different , and really good.
Thanks very much George. Pretty great stuff.
One note for other potential builders. I used pots I had on hand. Single turn like those on the BOM but different make. The ones I used are twitchy like stink, going from 40mA to 80 with just thinking about turning the screwdriver. It likely would have been better to get the one George recommended. A multi-turn pot kluged to fit the pads might make the dialing in a little easier too.
6JC6 into 6CD6 - chosen because I'd seen them pictured in Tubelab's photos and figured following in the master's footsteps was the surest way to success without glitches.
Had a couple of minutes figuring out why the outputs weren't conducting until I realized the current sense 1Ω resistor pads for short and long bodies weren't joined. Then up and running without any trouble at all .
Input tube voltages are currently at 123Vdc and 6CD6 outputs are both fairly steady at around 55mA , wandering a little by about +/- 5mA.
B+ is roughly 480Vdc. Interesting that the output currents seem to be tracking together, maybe as line voltages fluctuate.
One comment : It only took a couple of minutes working on the board to get a feel for the years of experience that went into it. Even pad/hole size relationships made the placing and soldering . . . . I dunno, just different , and really good.
Thanks very much George. Pretty great stuff.
One note for other potential builders. I used pots I had on hand. Single turn like those on the BOM but different make. The ones I used are twitchy like stink, going from 40mA to 80 with just thinking about turning the screwdriver. It likely would have been better to get the one George recommended. A multi-turn pot kluged to fit the pads might make the dialing in a little easier too.
There is music! Congratulations.With a B+ of 480V you probably have 420v across the 6CD6, and at 55 ma that makes 23 watts dissipation (guess this includes screen current if you measure current at the 1 watt tail resistor). See any redness on the plates; spec for 6CD6 is 20 watts plate dissipation.
Could you tell more about what you hear? Any hum? And measure; power out and distortion? What output transformers did you use?
Last edited:
Hi Francois,
It runs quiet with the only hum coming from the AC filament of the linestage feeding it.
The quiescent current on the output tubes does seem to have issues from one powered session to the next. Last night after listening for a few minutes and hearing a problem I checked the output currents and one channel was down 10mA from 55 and the other was running at 4mA. Amazingly there was still bass centred between the two speakers !)
I think though that the pot is the problem with a sticky wiper that can spring a long way once it unglues itself.
I do know the practice of backing off the screw a touch after reaching the desired setting but it doesn't seem to relieve the tension enough to prevent the springback at some later time, I guess triggered by vibration. I've left my meter hooked up the the amp to use as a current meter until I get around to replacing the pots.
It also burps a bit on startup while the voltages are sorting themselves out. A couple of times it was about one second of motorboat in one channel.
(Actually, I called it motorboating but I don't think it's the power supply. A few years ago I built a couple of versions of Steve Bench's LRnoC phono stage with positive feedback in the cathode circuit [but without his muting subcircuit] and it sounds like that though much less pronounced and much much shorter duration.)
All other power-ups (and power -downs) have been just low level hums coming and going until it has sorted itself out and then steady and quiet.
No glow on the plate. I think it could go up even more without problems.
I am in the middle of sorting other things out and probably won't get around to any measurements for at least a week, though I'll try to if the time slot appears.
and just a note, Nothing I've said here is a complaint , not at all, (this is the best way to learn new stuff and I really like it) but I figure it's good to put all this out there so Tubelab can hear what builders are experiencing and adjust the circuit (or , more likely our level of knowledge : ) as needed.
It runs quiet with the only hum coming from the AC filament of the linestage feeding it.
The quiescent current on the output tubes does seem to have issues from one powered session to the next. Last night after listening for a few minutes and hearing a problem I checked the output currents and one channel was down 10mA from 55 and the other was running at 4mA. Amazingly there was still bass centred between the two speakers !)
I think though that the pot is the problem with a sticky wiper that can spring a long way once it unglues itself.
I do know the practice of backing off the screw a touch after reaching the desired setting but it doesn't seem to relieve the tension enough to prevent the springback at some later time, I guess triggered by vibration. I've left my meter hooked up the the amp to use as a current meter until I get around to replacing the pots.
It also burps a bit on startup while the voltages are sorting themselves out. A couple of times it was about one second of motorboat in one channel.
(Actually, I called it motorboating but I don't think it's the power supply. A few years ago I built a couple of versions of Steve Bench's LRnoC phono stage with positive feedback in the cathode circuit [but without his muting subcircuit] and it sounds like that though much less pronounced and much much shorter duration.)
All other power-ups (and power -downs) have been just low level hums coming and going until it has sorted itself out and then steady and quiet.
No glow on the plate. I think it could go up even more without problems.
I am in the middle of sorting other things out and probably won't get around to any measurements for at least a week, though I'll try to if the time slot appears.
and just a note, Nothing I've said here is a complaint , not at all, (this is the best way to learn new stuff and I really like it) but I figure it's good to put all this out there so Tubelab can hear what builders are experiencing and adjust the circuit (or , more likely our level of knowledge : ) as needed.
Spent some time doing more measurements with REW. For this test I plugged in a set of "winged" 6DQ5 outputs with the assumption that these could take some abuse along with 6JEJ7 drivers. My focus was to find the lowest THD at 1W. So with the DMM showing 2.83V at the speaker leads I started adjusting the pots.
Best THD was .58% but this was with plate voltage of 65 volts and bias at 105 mA. Also when I turned up the volume output would get to around 7-8W and then the amp would go unstable and need to be shut off to recover.
Best THD was .58% but this was with plate voltage of 65 volts and bias at 105 mA. Also when I turned up the volume output would get to around 7-8W and then the amp would go unstable and need to be shut off to recover.
More testing with REW. I tried same test at 1W with the 6KT6 for drivers. Similar results but THD would never go below 1.1%. Next I plugged in the 6J51P tubes. Results were the same as the 6JE7 tubes.
Now I wanted to see what would happen if I tried to adjust for lowest THD but at a higher output. So leaving the 6J51P tubes in and resetting the plate voltage to 125V and the bias to 80mA I cranked the source up so the amp was putting out 6W and spun the pots. First lowering the plate voltage and raising the bias lowered the THD as in the 1W test but then when the plate voltage went below about 95V the THD begin to rise again. So I started turning it the other way. THD started to go up then after around 140V started to go down until I got past 165V.
Now with the settings at 160V plate voltage for the driver and 100mA bias for the output tube THD at 1W was 1% and 5% THD came at 12W. Nothing was turning red after about 10 minutes of testing but again this was with the "winged" 6DQ5 output tubes which there has been some speculation that they can take a bit more than the average 6DQ5.
Now I wanted to see what would happen if I tried to adjust for lowest THD but at a higher output. So leaving the 6J51P tubes in and resetting the plate voltage to 125V and the bias to 80mA I cranked the source up so the amp was putting out 6W and spun the pots. First lowering the plate voltage and raising the bias lowered the THD as in the 1W test but then when the plate voltage went below about 95V the THD begin to rise again. So I started turning it the other way. THD started to go up then after around 140V started to go down until I got past 165V.
Now with the settings at 160V plate voltage for the driver and 100mA bias for the output tube THD at 1W was 1% and 5% THD came at 12W. Nothing was turning red after about 10 minutes of testing but again this was with the "winged" 6DQ5 output tubes which there has been some speculation that they can take a bit more than the average 6DQ5.
Just make sure we get this correctly - your B+ measured at r1 is still at 375V? 65 V is the driver plate voltage and the 105 ma is the cathode current of the 6DQ5s via Mosfet?
Correct. I don't expect to run the amp long term like this but it was an interesting test so I thought I would share. Also like I mentioned the amp would fall apart around 6-7W at these settings.
The charts that George published predict something closer to 14W with 80 mA of cathode current using the 3K OPTs and a B+ between 350-400V. That along with well under 1% THD at 1W is what I am hoping to get close to.
I believe that you put a 2K ohm resistor in series with the mosfet in the screen supply. You may need to lower it's value some.
I have seen the amp go unstable like this when driven to hard clipping. Put a voltmeter on the screen grid supply to the output tubes. Crank up the drive slowly and watch the screen voltage. If it starts to drop as you approach clipping the screen supply you need to lower the resistor value some. The ideal resistor size varies with the choice of output tube, and how well they were manufactured. You would want the value low enough to protect the mosfet and the screen grids in the output tubes, but not too low as to affect the amps operation.
If I set the bias at 75mA I can see the screen voltage start to sag at 12W. Moving the bias down to 65mA it goes unstable around 6-7W with no sag in the screen voltage. Even so I tried paralleling another 2K resistor to adjust the value down to 1K. No sag at 75 mA and 12W but it still went unstable at 65 mA.
Not sure what to try next to get closer to 14W at 5% THD and 0.4-0.7% at 1W. With the output tubes at 80 mA bias I can just get it to around 1% THD at 1W and 9W at 5%. To summarize my current setup B+ is at 375V and I have the 3K Toroidy OPT driving 8 ohm resistors from the 8 ohm tap. Output tubes are 6DQ5 and driver tubes are 6EJ7. Any ideas would be appreciated.
A few more updates since I had a little time to play this evening and was curious about a some other tests that came to mind. First I reset my baseline and with 80mA bias and 80V for the drivers I ran REW. About 1% THD at 1W and 11W at 5% with the 6DQ5 and 6EJ7 combo. 11W is a bit better than I remembered earlier.
I wanted to understand a bit more the effects of the resistor on the mosfet for the screen supply. With 2K in place the screen voltage stays solid at 190V till around 11-12W where it starts to drop fairly rapidly. Bringing the resistor value down doesn't really impact the output, even with no resistor and a steady 190V the amp will only put out a small fraction of a watt more before it exceeds 5% THD.
Next I was curious how different loads would impact the amp. First I hooked up a 4 ohm load to the 8 ohm OPT tap. THD went way up so I abandoned this and instead hooked a 8 ohm load to the 4 ohm tap. THD at 1W was now much lower, I was seeing .38%, however 5% came at only 5-6W.
After this I hooked it back up to some speakers and let it play. For the next few days it will be my garage amp as I work on other things over the holiday week.
I wanted to understand a bit more the effects of the resistor on the mosfet for the screen supply. With 2K in place the screen voltage stays solid at 190V till around 11-12W where it starts to drop fairly rapidly. Bringing the resistor value down doesn't really impact the output, even with no resistor and a steady 190V the amp will only put out a small fraction of a watt more before it exceeds 5% THD.
Next I was curious how different loads would impact the amp. First I hooked up a 4 ohm load to the 8 ohm OPT tap. THD went way up so I abandoned this and instead hooked a 8 ohm load to the 4 ohm tap. THD at 1W was now much lower, I was seeing .38%, however 5% came at only 5-6W.
After this I hooked it back up to some speakers and let it play. For the next few days it will be my garage amp as I work on other things over the holiday week.
I’ve been thinking about the sudden onset of the drop in the 190Vdc screen supply voltage at 11-12 watt that you observed. And it even happens without “2000 ohm resistor”.
Why would that be? Is R4’s value at 75k too high and starves D3 when the B+ drops below a certain threshold on high demand? With your B+ at 375V at no-load it appears that Izt is only ~2.6 ma, where the 1N5386B is specified Izt=5 ma.
I hope we could figure this out as well as the ~1% distortion at 1 watt. The UNSET could not be called a Hifi amp with those distortion figures at 1 watt.
To clarify without the 2K resistor the screen voltage does not drop. What remains unchanged regardless of the resistor value is the 11W limit in output at a measured 5% THD. Without the resistor I can keep turning it up and the watts and THD keeps rising. With the resistor the watts output just hit a wall and only the THD rises and the screen voltage drops.
I also picked up some 12GN7’s to experiment with for the driver tubes. Results are the same as with the 6EJ7 or 6J51P tubes.
One more observation that I touched on before that I think is relevant. I can get the THD down in the 0.4-0.7% range at 1 watt by adjusting the driver plate voltage to 55-65V. However the decrease in distortion is primarily coming from a reduction in the 2nd harmonic to the point that the amp becomes 3rd order dominant. This also brings down the output to 9W.
Subjectively I find the amp sounds best set around 110V at the driver plate. I think the real goal would be to figure out how to reduce the THD proportionately or only in the odd orders for that classic SET profile.
Honestly though with my tin ears I would have been happy with the sound had I not gone down the measurement rabbit hole but hey that is part of the fun and I did specifically jump into this beta build to learn and hopefully provide some useful data. I am sure it is clear I am starting from a zero knowledge base. Truth is I am an accountant by trade so this is far outside my wheelhouse.
From my perspective I would really like to see how other people's builds measure and the results George gets when he tries to duplicate my setup. I figure there is a range of possible reasons I am not seeing what was predicted. The most likely in my head as always are the ones that involve a mistake on my part
These beingAssembly error
Faulty/wrong parts
My combination of tubes and transformers
Testing setup issues
Errors in the build documentation
Tweaks needed in the design
Errors in the predicted results
As a side note I have tried 3 different brands of 6DQ5 output tubes; Sylvania, GE, and RCA. Each of these had clearly different construction. The GE's look like the majority of the other tubes I have, the Sylvania's have the extra "wings" added, and the RCA's were taller than everything else. The GE's and Sylvania's performed the same at 80mA bias and both have played for days at a time without issue. I also pushed the Sylvania's hardest at up to 105mA bias for short time periods and they never complained. The RCA's gave slightly more THD across the board than the other tubes and after about an hour of operation it all ended with the sounds of loud pops from the speakers and a light show in one of the tubes.
Hehe! I thought the “it” that starts to drop was the screen voltage. You meant the output. Sorry about my misunderstanding.
So, your screen voltage is rock solid at 190Vdc regardless of the output level?