A while back I remember Pete Millett was experimenting with heating tube filaments at an AC frequency well above human hearing. I thought that was a very interesting approach. I was thinking, (dangerous newby), couldn't I do that with a cheap $12 class D amp board and a $5 Chinese sine wave oscillator board. I have some small ($12 each) class D boards that actually put out 60 and 120 watts to a speaker, isn't that enough to also heat a vacuum tube? Can I just play the oscillator at lets say 80kHz into the input, then hook the tube to the speaker terminals, then set the voltage using the volume control? To make a high frequency AC heater supply.
Take a look at Alex Kitic's solution for AC heating. Go about a third of the way down the page.
https://rh-amps.blogspot.com/2014/06/rh813-flagship-project.html
I tried it with my 6SK17-v / GK71 amp where I was trying to build a lightweight transmitting tube amp. It worked well for me and my old ears, easily the best amp I have ever built.
https://audioratbag.blogspot.com/2016/01/the-heart-attack-special-6s17k-v.html
I disassembled the amp when I moved to Australia in 2020 and am currently rebuilding it but this time I will try some different DC solutions. The first will be an LT1083 version of the D-Noizator from the Power Supply forum.
https://www.diyaudio.com/community/...-retrofit-upgrade-any-317-based-v-reg.331491/
ray
https://rh-amps.blogspot.com/2014/06/rh813-flagship-project.html
I tried it with my 6SK17-v / GK71 amp where I was trying to build a lightweight transmitting tube amp. It worked well for me and my old ears, easily the best amp I have ever built.
https://audioratbag.blogspot.com/2016/01/the-heart-attack-special-6s17k-v.html
I disassembled the amp when I moved to Australia in 2020 and am currently rebuilding it but this time I will try some different DC solutions. The first will be an LT1083 version of the D-Noizator from the Power Supply forum.
https://www.diyaudio.com/community/...-retrofit-upgrade-any-317-based-v-reg.331491/
ray
I mentioned the Pete Millett high frequency AC filament heating article but neglected to go find it earlier:
http://pmillett.com/hf_fil.htm
Its what started me thinking about how to make a high power / high frequency AC supply at all.
http://pmillett.com/hf_fil.htm
Its what started me thinking about how to make a high power / high frequency AC supply at all.
I did some experiments along these lines a long time ago.....before Class D amps existed, and before Tubelab and the TSE existed. I learned that the usual AC filaments on a DHT like a 45 or 300B will generate low level Inter Modulation Distortion tones (mixing produces) even if the related filament hum is nulled well into the inaudible range. These IMD products make the amp sound "fuller" to some people, and cause listener fatigue in others. I am from the latter group.
I did successfully get a 45 tube to run from a vacuum tube powered AC source that could produce a few watts at 100 KHz. A HP 200CD wide range oscillator has a push pull pair of 6CW5 tubes capable of making 5 to 10 watts, an air core OPT for the higher frequency ranges, and goes to 600 KHz. The low impedance output feeds a pair of resistors to get to 600 ohms. I bypassed the resistors and the output attenuator, and barely got enough power to light a single 45 tube. Here the filament "hum" and all of the IMD products that are generated by the heater voltage intermodulating with the audio are all way above the audio frequency range and will never get through the amps OPT. The only distortion products seen in the amps output are harmonics.
I considered building such a system and even did some crude experiments, but the chip amps available in the late 1990s will not work at 100 KHz or get so hot that they self destruct in a few minutes. It was probably possible to build such a system with discrete semiconductors in the late 90's, but it would be too large and complicated for what became the Tubelab TSE board which uses DC heating for all four of the tubes.
You would need a relatively low distortion sine source, a linear amp of some sort, a feedback or other control loop to keep the output voltage constant, and a DIY output transformer. I have thought about it several times since then, but never actually built anything.
I did successfully get a 45 tube to run from a vacuum tube powered AC source that could produce a few watts at 100 KHz. A HP 200CD wide range oscillator has a push pull pair of 6CW5 tubes capable of making 5 to 10 watts, an air core OPT for the higher frequency ranges, and goes to 600 KHz. The low impedance output feeds a pair of resistors to get to 600 ohms. I bypassed the resistors and the output attenuator, and barely got enough power to light a single 45 tube. Here the filament "hum" and all of the IMD products that are generated by the heater voltage intermodulating with the audio are all way above the audio frequency range and will never get through the amps OPT. The only distortion products seen in the amps output are harmonics.
I considered building such a system and even did some crude experiments, but the chip amps available in the late 1990s will not work at 100 KHz or get so hot that they self destruct in a few minutes. It was probably possible to build such a system with discrete semiconductors in the late 90's, but it would be too large and complicated for what became the Tubelab TSE board which uses DC heating for all four of the tubes.
You would need a relatively low distortion sine source, a linear amp of some sort, a feedback or other control loop to keep the output voltage constant, and a DIY output transformer. I have thought about it several times since then, but never actually built anything.
I wonder what problem a HF filament supply solves that a DC filament doesn't solve more effectively. The modern DC/DC switchers operate either just below or just above the AM radio band and use spread-spectrum clocking or some other method of dithering to avoid interference. Any ripple they may have is well above human hearing.
I first started using the DC/DC switching modules and ICs some 13 years ago. You can read about them here: https://www.diyaudio.com/community/threads/universal-filament-regulator.189693/ with a significant update here: https://www.diyaudio.com/community/threads/universal-filament-regulator.189693/post-7580598. They worked wonderfully with 300B tubes and many others.
Tom
I first started using the DC/DC switching modules and ICs some 13 years ago. You can read about them here: https://www.diyaudio.com/community/threads/universal-filament-regulator.189693/ with a significant update here: https://www.diyaudio.com/community/threads/universal-filament-regulator.189693/post-7580598. They worked wonderfully with 300B tubes and many others.
Tom
Oh? How so?
As far as I understand it, the cathode/filament will emit electrons once it gets hot enough. These electrons from a space charge cloud around the cathode as they repel each other. As all electrons have the same charge, they will be evenly distributed in the space charge cloud and one electron will be indistinguishable from the other. Why would it matter whether the cathode is heated by LF AC, HF AC, or DC ... or by a heater?
Tom
I used to concerned about that article from EML too. It isn't totally bs, but it is close to it.
How do we know that DC is better than high frequency AC? Because there is no regulator, power supply, or voodoo magic that will ever sound better than a battery. A couple D cell batteries and a trim resistor is incredibly hard, if not impossible, to beat.
In an indirectly heated tube it makes no difference how the cathode is heated except for the possibility of coupled AC causing hum.
In a directly heated tube, say one with a 5 volt heater with one end grounded and the other end fed +5 volts, having a negative 20 volts on the grid, we have a voltage gradient along that heater wire of 5 volts. One end will have 20 volts of negative bias WRT G1, and the other end will have 25 volts of negative bias WRT G1, Due to the differences in voltage across the heater, there will be more tube current flow at the end of the grounded end of the heater as there is less negative bias than the end fed with +5 volts.
In a directly heated tube, say one with a 5 volt heater with one end grounded and the other end fed +5 volts, having a negative 20 volts on the grid, we have a voltage gradient along that heater wire of 5 volts. One end will have 20 volts of negative bias WRT G1, and the other end will have 25 volts of negative bias WRT G1, Due to the differences in voltage across the heater, there will be more tube current flow at the end of the grounded end of the heater as there is less negative bias than the end fed with +5 volts.
That doesn't make sense to me, though. As far as I understand it, the current doesn't flow directly from the cathode. Rather the electrons are attracted from the space charge cloud by the anode of the tube. Once the electron has joined the cloud there's no way to know which end of the filament/cathode it came from.
Now, it may be that the current density is higher in areas that have lower Vgk, but I don't think that necessarily translates to a location on the cathode itself. All that is baked into the characteristic curves of the tubes as far as I understand it. Or am I off track here?
If the characteristics are dramatically different between DC and AC heating it should easy to test that theory by measuring the characteristic curves for the two types of filament supplies. Have you ever measured this? If there is a difference I would love to see hard data that shows how much the difference is. Comparing against common tube-to-tube variation would be an added bonus.
I agree with respect to the indirectly heated tubes, though.
Tom
Now, it may be that the current density is higher in areas that have lower Vgk, but I don't think that necessarily translates to a location on the cathode itself. All that is baked into the characteristic curves of the tubes as far as I understand it. Or am I off track here?
If the characteristics are dramatically different between DC and AC heating it should easy to test that theory by measuring the characteristic curves for the two types of filament supplies. Have you ever measured this? If there is a difference I would love to see hard data that shows how much the difference is. Comparing against common tube-to-tube variation would be an added bonus.
I agree with respect to the indirectly heated tubes, though.
Tom
No reason but the intrigue of it, and the serendipitous notion of an audio amplifier heating a tube or making your toast.
I struggled for weeks thinking... "how does someone actually build an AC power supply that can both provide a lot of power and at any desired frequency?". I don't possess the knowledge to begin to do that. But last night in bed it hit me, like things often do... When I put a dummy load on an amplifier, the resistor gets hot. Thats power. A speaker is a motor. Thats power. And a wide bandwidth amplifier can amplify a high frequency sine wave. So can this be the makings of an "on the cheap" HF AC power supply?
In a typical amp, the preamp and PI stage tubes typically operate with cathode resistors that would normally have a low heater-induced hum voltage across them because the heater-cathode parasitic resistance is many many megohm, and the parasitic capacitance is in the low pF, and cathode bypassing helps. That residual hum gets amplified on its way to the speakers, although is commonly suppressed to be just or below audibility. You may find a few pF at tens to hundreds of kHz starts to introduce some HF ripple into the audio chain, and it may have gain along the way to your output stage.
That's not true. It comes from wrong interpretation of how a vacuum tube works. If it were true one could often see hot spots on the plate which doesn't happen. I have never seen such a thing. And modern manufacturers of boutique tubes simply do NOT have any reasonable statistics supporting their claims with facts. In fact, EML tubes work great and better with DC heating! Actually if you buy KR equivalent tubes they WILL recommend DC heating....
I remember someone, maybe Rod Coleman, posting a few years ago a manufacturer's (RCA or similar) statistics about transmission DHTs tube life with DC and AC heating. DC was definitely better just because it can be regulated without many troubles. There is less distortion too (i.e. no intermodulation with PSU if properly done) which is the main reason why I use it.
Last edited:
I think so as well. In fact the very linear 45 tube curves in the datasheet are with DC heating and they look the same with AC.....
https://frank.pocnet.net/sheets/084/4/45.pdf
Back in the late 90's. John Atwood( One Electron fame) designed an amplifier using R.F. to heat the filaments. Hoe found there were
Advantages and dis-advantages. His big reason for doing R.F. for filaments is to get the AC way out of the ballpark as far as causing hum
and such. it did work but, the big issue he had to deal with was matching his R.F. generator to the effective load of the filament chain. I do not remember the frequency he used but, it was in one of the Amateur Radio bands as I remember. Here is the issue he ran into. The typical radio transmitter is designed to operate into a 50 Ohm load. A simple design example. A stereo SE EL34 amp using a 6SN7GT driver. All the filaments are basically in parallel. The EL34's have a filament resistance of around 4.2 ohms and the 6SN7's are around 20 ohms ea. When they are all in parallel, the resistance is 1.76 ohms real. The matching circuit becomes a bit touchy at this point. In order to get good power transfer, you have to set up a network with a very high Q and low loss in the component because it still has to deliver 4.2 Amps of current to light the tubes. He did do it but, it wasn't easy. Another advantage to the RF heated filaments. It drops the over all noise floor of the amp another 13 to 17 dB. You could walk into the test room and the only way you knew the amp was of was if it had a pilot lamp. If you had the input up too high when you started the music, you would go from virtually dead silence to wall shaking!!! Why High Frequency AC vs DC filaments. There is an issue with DC filaments in that unlike AC the metal in the filament stays put because of the constant changing from + to - . When you DC your filaments, you are actually shortening the Life of the tube because there is a constant differential between the 2 terminals of the filaments and the metal will migrate from one side to the other ultimately opening the filament eventually.
One other filament method that does work is one I adopted a long time ago. It wasn't original just, a bit obscure. You float the filament chain above AC ground. Then use 2 .01 caps one on each side of the filament chain. For all practical purposes you will now be in Hum Balance. Now you make a low current 10 to 1 resistive divider on the output of the power supply. If you HV supply is at 350 V then at the divider at the junction will be around 35V. Add a 100 uF cap at the 35 V supply point and run just a single wire from the 35 V supply point to 1 side of the filaments. With the 2 .01 caps, you are effectively at AC ground but you are floating from the DC HV supply. The hum will be non-existent and then the DC 35 V elevates the the filaments up above ground and it drops the overall noise floor of the amp at least 13 dB and usually more. So, there is 2 different methods of not using DC filaments. I used the method I described when I designed the power amps for Granite Audio in the late 90's but, I had been using it for years on my own HiFi amps and my guitar amps. THe elevated DC concept comes from the old tube radio manufactures. Philco, Zenith and such. I really like the elevated DC method but, I have wanted to try John's method because it does work.
Regards to all,
James
Advantages and dis-advantages. His big reason for doing R.F. for filaments is to get the AC way out of the ballpark as far as causing hum
and such. it did work but, the big issue he had to deal with was matching his R.F. generator to the effective load of the filament chain. I do not remember the frequency he used but, it was in one of the Amateur Radio bands as I remember. Here is the issue he ran into. The typical radio transmitter is designed to operate into a 50 Ohm load. A simple design example. A stereo SE EL34 amp using a 6SN7GT driver. All the filaments are basically in parallel. The EL34's have a filament resistance of around 4.2 ohms and the 6SN7's are around 20 ohms ea. When they are all in parallel, the resistance is 1.76 ohms real. The matching circuit becomes a bit touchy at this point. In order to get good power transfer, you have to set up a network with a very high Q and low loss in the component because it still has to deliver 4.2 Amps of current to light the tubes. He did do it but, it wasn't easy. Another advantage to the RF heated filaments. It drops the over all noise floor of the amp another 13 to 17 dB. You could walk into the test room and the only way you knew the amp was of was if it had a pilot lamp. If you had the input up too high when you started the music, you would go from virtually dead silence to wall shaking!!! Why High Frequency AC vs DC filaments. There is an issue with DC filaments in that unlike AC the metal in the filament stays put because of the constant changing from + to - . When you DC your filaments, you are actually shortening the Life of the tube because there is a constant differential between the 2 terminals of the filaments and the metal will migrate from one side to the other ultimately opening the filament eventually.
One other filament method that does work is one I adopted a long time ago. It wasn't original just, a bit obscure. You float the filament chain above AC ground. Then use 2 .01 caps one on each side of the filament chain. For all practical purposes you will now be in Hum Balance. Now you make a low current 10 to 1 resistive divider on the output of the power supply. If you HV supply is at 350 V then at the divider at the junction will be around 35V. Add a 100 uF cap at the 35 V supply point and run just a single wire from the 35 V supply point to 1 side of the filaments. With the 2 .01 caps, you are effectively at AC ground but you are floating from the DC HV supply. The hum will be non-existent and then the DC 35 V elevates the the filaments up above ground and it drops the overall noise floor of the amp at least 13 dB and usually more. So, there is 2 different methods of not using DC filaments. I used the method I described when I designed the power amps for Granite Audio in the late 90's but, I had been using it for years on my own HiFi amps and my guitar amps. THe elevated DC concept comes from the old tube radio manufactures. Philco, Zenith and such. I really like the elevated DC method but, I have wanted to try John's method because it does work.
Regards to all,
James