I have been working for sometime on a basic push-pull power amplifier using the Russian 6P13S sweep tube and the 6F12P pentode-triode.
Recently I started testing my working proto type so I thought I would post some results in case there is interest.
The 6F12P pentode-triode was selected at the time I started as it has a high transconductance of S=19 allowing high gain and good bandwidth. The triode section although limited in peak current also has the same S=19 allowing it to track closely the pentode signal when used as a split load phase inverter. The result is high gain, low THD and wide bandwidth.
Unfortunately they have become harder to find with a lots now on the market that are used and too low in emissions to be of any use. Pity as it is a good tube and for personal reasons I am unable to buy from russia at this point in history while hoping for better times in the future.
Sweep tubes are interesting beasts as even the smaller sizes like the 6P13S given appropriate screen voltages they can deliver high levels of peak plate current. Often far higher than for example a typical 6L6.
The high peak plate current provides some benefits on musical peaks at both the higher and lower frequencies where higher levels of peak current can be required due to transformer and loudspeaker issues. A lot of tube designs measure great to 1KHz with sine waves into the designed impedance but struggle to produce musical peaks into off impedance loads like a real loudspeaker. At frequency extremes transformer issues can also limit performance due to lack of available peak current from winding capacitance at high frequencies or magnetizing inductance current at low frequencies.
So I though to see what a higher peak current sweep tube design would do designed into a 6L6 class power range of about 30Watts RMS. I did not use the larger sweep tubes as they have become expensive and hard to get but the 6P13S is still low cost. Admittedly it would have been easier to just use the trusty 6L6 but where is the fun in that?
There is reason to believe sweep tubes can be pushed past their rating with some confidence as they were designed to run hot at full power and very high peak voltage and current continually with years long reliability in TV use. Some old application notes suggested 40% over the sweep rating for audio use.
Most sweep tubes will not tolerate the high screen voltages that the 6L6, EL34 or KT88 will making UL operation a bit more challenging. Custom transformers with a separate screen winding is one solution but far too costly for this project. I used a screen connection for the 6P13S UL operation with the AC portion of the screen signal capacitance coupled from the output transformer UL tap and the DC screen current feed from a regulated DC power supply. As the sweep tubes are more sensitive to screen voltage changes I further reduced the UL AC signal level by a two resistor voltage divider and this provided the lowest THD of all versions.
This simulated and worked well on the bench and compared well to straight pentode operation. UL operation seemed at first to provide little advantage over straight pentode operation in distortion or stability until I noted output peaking at 500Khz occurred in the pentode version when the output load impedance was mismatched to the transformer.
This suggested that dynamic stability would likely be improved with the local feedback that UL operation provides in the output stage.
Below is the spice circuit and spice THD test results. Note total THD is highest with 40% UL operation and lowest with the UL AC screen signal reduced by 4. Total power output was a bit lower in 40% UL. Square wave response was unchanged.
All simulated at 30watts output into 8 ohms.
Pentode operation, regulated screen THD=0.1625187 PERCENT
40% UL, regulated screen THD = 0.4871198 PERCENT
40% UL level divided by 4, regulated screen, THD = 0.1262827 PERCENT
With UL screen drive reduced from 40% by a factor of 4 distortion dropped and power levels returned to near pentode levels. This UL version seemed a good compromise in added complexity in exchange for potentially improved stability.
In spice UL 8 ohm frequency response was -0.3dB down at 20Hz and -0.24dB 20KHz and with peaking down -11dB at 1Mhz frequency measured into 8 ohms and at 8,000 ohms load down -4.5dB ohms at 1.3Mhz suggesting stability. Note we are looking at a audio amplifier's frequency and phase response up and into the Mhz. This has nothing to do with reproducing these frequencies but does give clues as the stability of the amplifier. You will note I look down to below .1Hz and well above 1Mhz in the plots. There is much going inside in a audio amplifier that happens far outside frequencies that audio test equipment can measure that is important in how a amplifier will sound when dealing with dynamic transients found in music. Spice makes seeing this easy but a lab setup to measure this on the bench is no easy task.
I created the spice models for the tubes using a Utracer 4+ to collect detailed data on the tubes and then Ronald's modeling software to create the spice models by measuring actual sample NOS tubes. Wonderful tool but there is a learning curve in creating models with this software. Highly recommended and well worth the effort.
Power output in spice was about 30 watts with THD at about 0.126 % and
bench testing showed that was not far off the real world with the proto type producing 26 watts at 0.094% THD at 2KHz and at 1 watt 0.0177% THD. The actually bench voltage rail being a bit lower than I used in spice could be why the power level is down a bit.
20Hz testing produced 26W @ 2%, 10W @ 0.478% and 1 watt @ 0.167%
200Hz at 1W @ was 0.0369%
20KHz produced 26W @ 1.3% and at 1 watt @ 0.198%
Square wave response in spice at 1Khz looked very clean with no ringing or over shoot and nice sharp edges.
Bench testing at 20Hz, 200Hz, 2KHz and 20KHz shows nice sharp clean edges confirming stable wide bandwidth.
The 20Hz square wave apart from the strong tilt due to low frequency loss shows monotonic lines suggesting a absence of saturation, strong non-linearity's or instability in the transformer and feedback loop a this very low frequency.
The 200Hz square wave has clean edges with modest tilt and no ringing or over shoot.
At 2Khz the square wave is very fast and clean with perhaps a tiny over shoot so small it could be a artifact of the scope probe setup (I did not take the time to calibrate the scope probe). 20Khz square waves are starting to show some loss of rise time but again are very well behaved with perhaps a small amount of leading edge over shoot. Considering this 20Khz waveform went through a great big hunk of copper and iron to me this is impressive and suggests the Hammond 1650T transformer used are very well designed and made.
If there is interest I will post more detail of the PCB layout and the actual PCB proto type amplifier circuit diagram as well as the regulated power supplies used for this project.
Recently I started testing my working proto type so I thought I would post some results in case there is interest.
The 6F12P pentode-triode was selected at the time I started as it has a high transconductance of S=19 allowing high gain and good bandwidth. The triode section although limited in peak current also has the same S=19 allowing it to track closely the pentode signal when used as a split load phase inverter. The result is high gain, low THD and wide bandwidth.
Unfortunately they have become harder to find with a lots now on the market that are used and too low in emissions to be of any use. Pity as it is a good tube and for personal reasons I am unable to buy from russia at this point in history while hoping for better times in the future.
Sweep tubes are interesting beasts as even the smaller sizes like the 6P13S given appropriate screen voltages they can deliver high levels of peak plate current. Often far higher than for example a typical 6L6.
The high peak plate current provides some benefits on musical peaks at both the higher and lower frequencies where higher levels of peak current can be required due to transformer and loudspeaker issues. A lot of tube designs measure great to 1KHz with sine waves into the designed impedance but struggle to produce musical peaks into off impedance loads like a real loudspeaker. At frequency extremes transformer issues can also limit performance due to lack of available peak current from winding capacitance at high frequencies or magnetizing inductance current at low frequencies.
So I though to see what a higher peak current sweep tube design would do designed into a 6L6 class power range of about 30Watts RMS. I did not use the larger sweep tubes as they have become expensive and hard to get but the 6P13S is still low cost. Admittedly it would have been easier to just use the trusty 6L6 but where is the fun in that?
There is reason to believe sweep tubes can be pushed past their rating with some confidence as they were designed to run hot at full power and very high peak voltage and current continually with years long reliability in TV use. Some old application notes suggested 40% over the sweep rating for audio use.
Most sweep tubes will not tolerate the high screen voltages that the 6L6, EL34 or KT88 will making UL operation a bit more challenging. Custom transformers with a separate screen winding is one solution but far too costly for this project. I used a screen connection for the 6P13S UL operation with the AC portion of the screen signal capacitance coupled from the output transformer UL tap and the DC screen current feed from a regulated DC power supply. As the sweep tubes are more sensitive to screen voltage changes I further reduced the UL AC signal level by a two resistor voltage divider and this provided the lowest THD of all versions.
This simulated and worked well on the bench and compared well to straight pentode operation. UL operation seemed at first to provide little advantage over straight pentode operation in distortion or stability until I noted output peaking at 500Khz occurred in the pentode version when the output load impedance was mismatched to the transformer.
This suggested that dynamic stability would likely be improved with the local feedback that UL operation provides in the output stage.
Below is the spice circuit and spice THD test results. Note total THD is highest with 40% UL operation and lowest with the UL AC screen signal reduced by 4. Total power output was a bit lower in 40% UL. Square wave response was unchanged.
All simulated at 30watts output into 8 ohms.
Pentode operation, regulated screen THD=0.1625187 PERCENT
40% UL, regulated screen THD = 0.4871198 PERCENT
40% UL level divided by 4, regulated screen, THD = 0.1262827 PERCENT
With UL screen drive reduced from 40% by a factor of 4 distortion dropped and power levels returned to near pentode levels. This UL version seemed a good compromise in added complexity in exchange for potentially improved stability.
In spice UL 8 ohm frequency response was -0.3dB down at 20Hz and -0.24dB 20KHz and with peaking down -11dB at 1Mhz frequency measured into 8 ohms and at 8,000 ohms load down -4.5dB ohms at 1.3Mhz suggesting stability. Note we are looking at a audio amplifier's frequency and phase response up and into the Mhz. This has nothing to do with reproducing these frequencies but does give clues as the stability of the amplifier. You will note I look down to below .1Hz and well above 1Mhz in the plots. There is much going inside in a audio amplifier that happens far outside frequencies that audio test equipment can measure that is important in how a amplifier will sound when dealing with dynamic transients found in music. Spice makes seeing this easy but a lab setup to measure this on the bench is no easy task.
I created the spice models for the tubes using a Utracer 4+ to collect detailed data on the tubes and then Ronald's modeling software to create the spice models by measuring actual sample NOS tubes. Wonderful tool but there is a learning curve in creating models with this software. Highly recommended and well worth the effort.
Power output in spice was about 30 watts with THD at about 0.126 % and
bench testing showed that was not far off the real world with the proto type producing 26 watts at 0.094% THD at 2KHz and at 1 watt 0.0177% THD. The actually bench voltage rail being a bit lower than I used in spice could be why the power level is down a bit.
20Hz testing produced 26W @ 2%, 10W @ 0.478% and 1 watt @ 0.167%
200Hz at 1W @ was 0.0369%
20KHz produced 26W @ 1.3% and at 1 watt @ 0.198%
Square wave response in spice at 1Khz looked very clean with no ringing or over shoot and nice sharp edges.
Bench testing at 20Hz, 200Hz, 2KHz and 20KHz shows nice sharp clean edges confirming stable wide bandwidth.
The 20Hz square wave apart from the strong tilt due to low frequency loss shows monotonic lines suggesting a absence of saturation, strong non-linearity's or instability in the transformer and feedback loop a this very low frequency.
The 200Hz square wave has clean edges with modest tilt and no ringing or over shoot.
At 2Khz the square wave is very fast and clean with perhaps a tiny over shoot so small it could be a artifact of the scope probe setup (I did not take the time to calibrate the scope probe). 20Khz square waves are starting to show some loss of rise time but again are very well behaved with perhaps a small amount of leading edge over shoot. Considering this 20Khz waveform went through a great big hunk of copper and iron to me this is impressive and suggests the Hammond 1650T transformer used are very well designed and made.
If there is interest I will post more detail of the PCB layout and the actual PCB proto type amplifier circuit diagram as well as the regulated power supplies used for this project.
Last edited:
Member
Joined 2009
Paid Member
You have a very nice test bench!
That’s a nice project. I share your restrictions on where I source my tubes these days too.
That’s a nice project. I share your restrictions on where I source my tubes these days too.
Very interesting approach to the “sensitive screen” problem of the ubiquitous TV tubes. Thank you for sharing and I am all ears to learn more of your project, especially the separated AC feedback and regulated DC supply for the screens of the output tubes. Best wishes and please keep us informed.
Interesting that you found 1/4 of the 40% UL AC feedback the lowest distortion. This approach could help us determine optimal UL feedback for various tubes without having access to transformers wound with different UL ratios.
Regarding the rigid regulation of the DC component of the screen supply, you may consider the approach Dave Gillespie has developed to modulate the screen voltage according to the B+ voltage he dubbed EFB II(tm). Apparently he found benefits in keeping the ratio between Screen DC and B+ constant, rather than variable when using a fixed regulated Screen DC supply and variable (due to sag) B+. This is explained in several threads on AudioKarma, for example this one:
https://audiokarma.org/forums/index.php?threads/improving-the-fisher-sa-100-with-efb-ii.476431/
Last edited:
If you decrease the 40 % UL tap signal to 1/4, doesn't this equal an UL tap of 10 %, signal wise? Including the µg1g2 parameter, I guess this would come very close to cathode feedback into the finals in true pentode mode, using the complete 16 Ω secondary with the 4 Ω tap grounded.
Best regards!
Best regards!
I stocked up on 6F12P because my amplifier designs all use it - there is no other tube quite like it. I can't even find another tube that uses the same pinout - so I have almost 200 of them left. I've never had to replace one even though I run them with a 600V B+
I use 6P13S strapped as triodes in the amp I'm listening to now. It makes almost 40WRMS @ 30Hz because I don't care about the datasheet figures - audio isn't nearly as demanding as sweep operation. I run them push pull, 320V B+, 2k2 Ra-a. They are awesome little tubes IMHO.
I use 6P13S strapped as triodes in the amp I'm listening to now. It makes almost 40WRMS @ 30Hz because I don't care about the datasheet figures - audio isn't nearly as demanding as sweep operation. I run them push pull, 320V B+, 2k2 Ra-a. They are awesome little tubes IMHO.
They are still pretty cheap - especially used ones (good as new) 42 pieces for 50$USD from Ukraine.
6F12P on eBay
6F12P on eBay
Re:"I use 6P13S strapped as triodes"
Using sweep tubes a triodes is a idea I will have to look into more.
The high sensitivity screen should allow the tube to deliver lots of current even at the low plate voltage at the bottom of the audio swing on the plate.
Thanks for the hint.
Using sweep tubes a triodes is a idea I will have to look into more.
The high sensitivity screen should allow the tube to deliver lots of current even at the low plate voltage at the bottom of the audio swing on the plate.
Thanks for the hint.
Apparently, I'm losing my mind - I'm using 6P31S in the aforementioned amplifier!
Here's the OP - Idle point is off the damned chart FFS!
Here's the OP - Idle point is off the damned chart FFS!
Yes, despite of what many eBay sellers tend to claim, 6P13S and 6P31S are different tubes. 6P31S comes close to EL36/6CM5.
Happy Xmas and a merry new year!
Happy Xmas and a merry new year!
@Bluesystems
You can also check Philips EV series of amps which uses EL36 (6CM5) tubes , IMHO Philips designed those amps in the right way , where EL36 are pentode connected with B+=340V and g2 =340V/2=170V , where that g2=170V are derived in one very simple and elegant way , always from power transformer secondary center tap .http://www.amplifiers-with-valves.nl/ev44xx.html
You can also check Philips EV series of amps which uses EL36 (6CM5) tubes , IMHO Philips designed those amps in the right way , where EL36 are pentode connected with B+=340V and g2 =340V/2=170V , where that g2=170V are derived in one very simple and elegant way , always from power transformer secondary center tap .http://www.amplifiers-with-valves.nl/ev44xx.html