I have been dinking around with filament voltage (dc) on both my 841 driver and 801a output tubes. Yes, I am aware of the starved filament adverse impact to tube life. But, and this is really BIG, without a doubt (subjectively), the sound is much much better with a lower filament voltage. Yes, I have read Steve Bench's finding regarding lower distortion as a result of starved filaments. But, could that really explain improved bass, deeper bass, deeper soundstage, more transparency? Transparency and soundstage improvements are logical. But, much better bass? So, then I started to wonder about the Rp relationship with filament voltage. Bench and other state that Rp actually increases with starved filament voltage. So, the higher Rp (caused by the starved filament voltage) is counter intuitive to my experience. Other benefits of starved filaments is a more constant mu and gm.
So, any thoughts on why the sound is so much better. BTW: Initially, the driver was starved and improvements were clearly noticeable. Then, I said, why not also starve the output tube. This again gave further improvement, most noticeably in the bass / lower registers.
Schematic is below. Yes, I know I have a mismatch between the 841 Rp (20k) and the 68H IT. And for the record, I have confirmed that the 841 loves (requires) higher plate voltage. I am running at 535V. Bela suggests even higher at 600V.
Finally, I have inserted Rod Coleman's latest V9 filament regs. FANTASTIC. Eliminate all noise with the prior regs and also improved the transparency & soundstage. I need to work on bass sound pressure, but that is likely a function of the IT / driver Rp mismatch.
So, thoughts on the filament starvation?
So, any thoughts on why the sound is so much better. BTW: Initially, the driver was starved and improvements were clearly noticeable. Then, I said, why not also starve the output tube. This again gave further improvement, most noticeably in the bass / lower registers.
Schematic is below. Yes, I know I have a mismatch between the 841 Rp (20k) and the 68H IT. And for the record, I have confirmed that the 841 loves (requires) higher plate voltage. I am running at 535V. Bela suggests even higher at 600V.
Finally, I have inserted Rod Coleman's latest V9 filament regs. FANTASTIC. Eliminate all noise with the prior regs and also improved the transparency & soundstage. I need to work on bass sound pressure, but that is likely a function of the IT / driver Rp mismatch.
So, thoughts on the filament starvation?
Attachments
Because all electron needs for power. 🙂
... when you want to use other anode load, than inductor.
I am not following. Please explain what you mean by "when you want to use other anode load, than inductor"???
Unless backed by measured performance changes: expectation bias.
JMF - I can't deny what you are saying. I'm just passing along what I have heard. Just one man's subjective observations (oxymoron?).
Hey, Pat!
801A and 841 are both thoriated tungsten filaments. This type can be run with reduced power, and the lifetime will be improved, not degraded!
It's a completely different matter with coated filaments, where underheating can be life-shortening if you take it too far.
It's worth checking that the gain and power output are unchanged when you reduce the heating power, but 5% low is a good start.
801A and 841 are both thoriated tungsten filaments. This type can be run with reduced power, and the lifetime will be improved, not degraded!
It's a completely different matter with coated filaments, where underheating can be life-shortening if you take it too far.
It's worth checking that the gain and power output are unchanged when you reduce the heating power, but 5% low is a good start.
The "enemy" of thoriated tungsten filament tubes is the on/off cycles, due to the Miller-Larson Effect.
Jerry C. Whitaker
Power Vacuum Tubes: Handbook
Morgan Jones :
https://books.google.hu/books?id=ed...gsten Filament "Miller-Larson Effect"&f=false
There are two reasons why the off/on cycle kills thoriated tungsten filaments:
https://www.audioasylum.com/cgi/t.mpl?f=tubediy&m=258874
Jerry C. Whitaker
Power Vacuum Tubes: Handbook
Morgan Jones :
https://books.google.hu/books?id=ed...gsten Filament "Miller-Larson Effect"&f=false
There are two reasons why the off/on cycle kills thoriated tungsten filaments:
- As the filament temperature passes through 900 degrees K [either when they are turned on or switched off], the Miller-Larson effect causes the grains of the metal to reorient themselves, so that the wire becomes thinner and longer. Worse, if a given section of the filament is slightly thinner, the increased current density causes increased localized heating, which exacerbates the Miller-Larson effect and causes further necking of the filament. Eventually, this necking leads to such deep cracks that the remaining conductive material has sufficiently high current density and local heating to vaporize it, thus destroying the filament.
- The resistance of a cold filament is far less than that of a hot one, and assuming an operating temperature of 1,975K, but an ambient temperature of 293K (20 deg. C), the initial cold current is 8.6 times higher than the operating current. The inrush current through the filament interacts with the Earth's magnetic field to produce a small kick. Combined with the Miller-Larson effect, this gradually deepens the surface cracks in the brittle filament. The damage done to the filament is proportional to the cube of inrush current, so a 'softstart' circuit can be worth while.
https://www.audioasylum.com/cgi/t.mpl?f=tubediy&m=258874
Last edited:
If you use 600V B+ and choke load, the anode voltage of tube near to the 600V.
At 600V the possible output swing is very limited in A1 region.
First off, as you already know, you are doing some weird stuff. 841 is a cool tube but I can't see how it would drive an IT well.
But to answer your question, the starved filaments sound good because they are producing an effect that you like. It could be that they are lowering distortion. More likely, they are adding some distortion in the bass frequencies and those harmonics are making the bass sound subjectively bigger, and adding other effects at higher frequencies that sound nice. I've played around with listening to various levels of distortion and that is what I have noticed as distortion goes up. You clean up the distortion and bass seems to suffer, but FR is improved when you measure. Some distortion can make plucked strings sound really nice, but when you start mixing lots of tones you get a lot of IM products that don't sound so nice. Lots of tradeoffs and the type of music you listen to is a major factor.
It'll be hard to get to the bottom of this for sure without doing an A/B measurement and seeing what is happening. Starving filaments will change the tube characteristics. It may make them better or worse. What happens in one situation may be the opposite of what happens in another.
But to answer your question, the starved filaments sound good because they are producing an effect that you like. It could be that they are lowering distortion. More likely, they are adding some distortion in the bass frequencies and those harmonics are making the bass sound subjectively bigger, and adding other effects at higher frequencies that sound nice. I've played around with listening to various levels of distortion and that is what I have noticed as distortion goes up. You clean up the distortion and bass seems to suffer, but FR is improved when you measure. Some distortion can make plucked strings sound really nice, but when you start mixing lots of tones you get a lot of IM products that don't sound so nice. Lots of tradeoffs and the type of music you listen to is a major factor.
It'll be hard to get to the bottom of this for sure without doing an A/B measurement and seeing what is happening. Starving filaments will change the tube characteristics. It may make them better or worse. What happens in one situation may be the opposite of what happens in another.
SpreadSpectrum - thanks for helping. Regarding the starved filament effect, I listen to chamber music (strings and piano). So, the added distortion actually might be producing a pleasant sonic, at least to me.
Yes, the 841 is less than optimal with an IT. So, I am building a new set of monoblocks to replace the 841. Likely a 3C24 or 35TG driver. Output = GM70. The 3C24, at lower voltages (say 400-450V) has an Rp approaching 10K (or somewhere near there). This is near the limit of a good IT's capability, but I am will to sacrifice the top and bottom. For example, I am happy with 30hz and 10khz -3db. So, now my challenge is finding a hi-end custom IT.
Yes, the 841 is less than optimal with an IT. So, I am building a new set of monoblocks to replace the 841. Likely a 3C24 or 35TG driver. Output = GM70. The 3C24, at lower voltages (say 400-450V) has an Rp approaching 10K (or somewhere near there). This is near the limit of a good IT's capability, but I am will to sacrifice the top and bottom. For example, I am happy with 30hz and 10khz -3db. So, now my challenge is finding a hi-end custom IT.
Very fun! I used the 841 as a high swing driver in my Unity-Coupled amp. It's pretty and distortion is nice and low. Unfortunately, the higher the plate idle voltage, the lower the distortion but also rp goes up, especially if you have to lower the idle current. Which is precisely what you don't want for the IT.
3C24, 35TG, and GM70 all sound fun to play with. I've never done a design with an IT myself but I always thought it would be fun since it makes the rest of the amp so simple. Of course, a good one costs $$$, but I think good iron is worth the investment.
3C24, 35TG, and GM70 all sound fun to play with. I've never done a design with an IT myself but I always thought it would be fun since it makes the rest of the amp so simple. Of course, a good one costs $$$, but I think good iron is worth the investment.
OK, Bela. You win 🙂 But here is the challenge: we need a inter-stage tranny alternative for the 841 driver or a replacement for the 841. My requirements: Thoriated Tungsten. mu > 25. IT or DC. No RC. No sand, just pure iron. 30-10khz -3db
Recently I discovered the 3C24 and 35TG which have about 50% lower Rp than the 841. This is a step in the right direction.
Your turn.
Recently I discovered the 3C24 and 35TG which have about 50% lower Rp than the 841. This is a step in the right direction.
Your turn.
Maybe ... but you want to use such -50W power- tubes, which ones even requires more power to drive on a grid, that all power amp generates.
These kV hungry tubes working well .... if the plate colour is dark/bright red. 😛
Why not try -as I wrote you earlier- 1:2 SUT->EML20->IT-> fix biased 801a->5k:8 topology?
At 2V peek input you get about 5W 4%THD.
Pat: one other check: when you reduced the filament heating of the 841, did you compensate for the bias?
The driver DHT is connected with Filament Bias. If you turn down the current-adjustment of the V9 filament regulator to get a lower filament voltage, the bias voltage decreases in turn. This means more anode-current, and that will often sound better, since the curves are more regular at higher current.
If you want to reduce the filament power, while keeping the bias the same as before, you must increase the value of the filament bias resistor.
OTOH, if you want to try the effect of increased anode-current, while keeping filament power the same, simply add a parallel resistor to the filament bias resistor. For example: you have 3Ω now - try adding 47Ω or 33Ω in parallel, and see if that's what makes the sound improve more.
The driver DHT is connected with Filament Bias. If you turn down the current-adjustment of the V9 filament regulator to get a lower filament voltage, the bias voltage decreases in turn. This means more anode-current, and that will often sound better, since the curves are more regular at higher current.
If you want to reduce the filament power, while keeping the bias the same as before, you must increase the value of the filament bias resistor.
OTOH, if you want to try the effect of increased anode-current, while keeping filament power the same, simply add a parallel resistor to the filament bias resistor. For example: you have 3Ω now - try adding 47Ω or 33Ω in parallel, and see if that's what makes the sound improve more.
Shocking to learn: Coleman is on to something. 😉
I put the driver tube filament voltage back to 7.5Vdc. Then started changing values for the filament bias resistor which changed the filament bias which changed the anode current. This had the same effect as changing the filament voltage. As the filament bias was reduced and therefore more anode current, the soundstage, attack and transparency improved.
Nice input, Rod. Much appreciated.
I put the driver tube filament voltage back to 7.5Vdc. Then started changing values for the filament bias resistor which changed the filament bias which changed the anode current. This had the same effect as changing the filament voltage. As the filament bias was reduced and therefore more anode current, the soundstage, attack and transparency improved.
Nice input, Rod. Much appreciated.
I am assembling my V-9 regulators now. Glad you like what you are hearing! Got any pictures that might show how you incorporated the regs into your amp?
Unfortunately, I just had a 3C24 fail due to an open filament. Tube was only a few weeks/months old. Hopefully, this was just a fluke!