New DHT heater

[fdegrove]

Hi,

An excellent idea, Guido.

This should be a major help to all those people wondering how to tackle DHT heating properly.
Your solution to the problem is what it should be IMO, congrats.

So far I prefer having the + side of the heater to ground, but there still is many ways to experiment

Curious to see technical suggestions as to why the grounded + side would be prefered.
Even though I heard of people claiming heating IDHT small signal valves with a negative voltage yields a small but audible improvement before, I never found a plausible explanation as to why this would be.

Looking forward to the 2.5 and 10V versions to please the numerous 2A3 fans and 845/211 freaks out there....

Cheers,

[Guido Tent]

Originally posted by fdegrove
Hi,

An excellent idea, Guido.

This should be a major help to all those people wondering how to tackle DHT heating properly.
Your solution to the problem is what it should be IMO, congrats.

Curious to see technical suggestions as to why the grounded + side would be prefered.
Even though I heard of people claiming heating IDHT small signal valves with a negative voltage yields a small but audible improvement before, I never found a plausible explanation as to why this would be.

Looking forward to the 2.5 and 10V versions to please the numerous 2A3 fans and 845/211 freaks out there....

Cheers,

Hello Frank,

Thanks for positive feedback !

On the "where to ground the tube" discusion, may hours will be spend. Anyhow, it is up to the users of what they find sounds best.

Working on the 2.5V version, bit tricky due the low voltages.

The 10V 3.25A version will arive soon !

cheers

[ThorstenL]

Konnichiwa,

Originally posted by fdegrove
Curious to see technical suggestions as to why the grounded + side would be prefered.

I would like an explanation too. I first came across "positive (AC) grounded" in the Verdier and le Maison 300B Amp (the circuit was the basis for my "legacy" Amp). I tried it the other way and positive AC grounded sounded better, DESPITE the voltage reg obviously shortcircuiting any AC developed across the heater....

Sayonara

[Guido Tent]

Originally posted by Kuei Yang Wang
Konnichiwa,



I would like an explanation too. I first came across "positive (AC) grounded" in the Verdier and le Maison 300B Amp (the circuit was the basis for my "legacy" Amp). I tried it the other way and positive AC grounded sounded better, DESPITE the voltage reg obviously shortcircuiting any AC developed across the heater....

Sayonara

Hi Sayonara,

In the case of DC supply to the heater, the majority of audio-current runs through the side with highest gm, e.g the side when + is tied to ground (or cathode resistor).

With AC on the heater, I never did the experiment as described (AC grounding). Should do it one day though, wonder if it matches with my preference in the DC situation.

The physical current distribution through the tube also depends on the geometry of the filaments. This may also play a role.

I do not have a single answer yet......

regards

[Rod Coleman]

I too have been experimenting with filament supply buffering recently. I had found with ac-heating that the type of trafo had a big influence - toroids sounded shrieky and aggressive whereas 1,3A wallknobs sounded muddy and distant. I guessed this was due to poor dielectrics in the trafo, and resolved to build a dc supply with audiofrequency buffering on BOTH filament terminals. The outcome was the most wonderful upgrade I have achieved in five years with 300Bs. Not only that, but all the parts are junk box or radio-rally stock, and best of all, the biggest improvement is worked on the cheapest 300Bs - eg: 1998 vintage Sovteks which you will have put away long ago if you have ac-heating. Gone is the grimy, murky, & noisy sound - now they project into the room with vigour and vivacity.

I briefly considered a complementary current source to isolate both filament terminals, but soon realised a much better circuit with current sinking on the negative, and a gyrator (aka electronic choke) to reject ripple on the positive supply. The schematic tells the tale.



The first (mains) trafo has its output centre tap grounded (300B ground) to eliminate dc and common-mode mains noise. It also serves to keep diode noise and current pulses from feeding back up the mains & into other equipment. Thanks (again) are due to Lynn Olson for voicing this idea.

The second (toroid) trafo is fullwave schottky rectified to minimise the number of clanking diodes compared to a bridge. The target output is 8,7 to 9,5V with 1V of ripple. The high ripple voltage is mandatory to increase the conduction angle and reduce peak current. Ripple is cut to the mV level by the action of the gyrator Q1, T1 and associated passives. The NPN-PNP complementary feedback pair (CFP) works somewhat like a darlington here but with the big difference that when faced with out-of-band signals (limit typically set by C13) the CFP looks like a high (=R3, 1,6K) impedance. So one can try different values for different amplifiers. C13 is also required for CFP stability since there is no output capacitor for the CFP (caps between the filament terminals or to the filament supply return will destroy the performance of this circuit BTW). Be sure you have a scope at hand if you mess with the CFP parts values/types. And for T1 try to get some Zetex FZT1051As from Farnell or Digikey - substitutes will degrade performance if the high Hfe over wide Ic is not available, together with low Cob. For Q1, choose a TO-220 part with Cob < 75pF and Ft > 30MHz. For Q1 and T2, use a 5 degC/W heatsink or better - and mount the sink in free air.

The filament return side is current regulated by T2, T3 and the sense resistor. This simple circuit answers perfectly the requirement for high audiofrequency impedance. Simply select the current sense resistor R4 to give 5,2V across your 300B filaments. 0,5 ohm is a starting value for 1,1A to 1,3A 300Bs, and you then add parallel values to home in. The voltage drops about 5% when warmed up, due to the tempco (-2mV/deg C for Vbe) of T3 - so mount this part in free air.

Do not use LM317 in this position if you want good performance. The dropout and ac impedance are markedly worse, it is noisy, and its thermal cutout preserves a tenpenny chip at the expense of your precious triode.

when set up correctly, there should be about 2V across each power transistor for a 9V supply. Don't let it get below about 1,6V for best ac-performance, and at the other extreme, take care not to use a raw dc supply much above 9V or you will roast the underside of your amp, and probably overstress the transistors. And If you are not comfortable about setting currents in your expensive triodes, or are unfamiliar with stability testing of transistors, or if you cannot confidently calculate and test the thermal design/heatsinking of power transistors, please don't attempt to build this circuit. It will be cheaper in the long run to buy Guido's module and avoid wrecking your DHTs.

[Ciscokid]

Hi Guido Tent,

Very nice board! Look forward to giving a pair a try later this year.

Regards,
Danny

[Guido Tent]

Originally posted by Rod Coleman
I too have been experimenting with filament supply buffering recently. I had found with ac-heating that the type of trafo had a big influence - toroids sounded shrieky and aggressive whereas 1,3A wallknobs sounded muddy and distant. I guessed this was due to poor dielectrics in the trafo, and resolved to build a dc supply with audiofrequency buffering on BOTH filament terminals. The outcome was the most wonderful upgrade I have achieved in five years with 300Bs. Not only that, but all the parts are junk box or radio-rally stock, and best of all, the biggest improvement is worked on the cheapest 300Bs - eg: 1998 vintage Sovteks which you will have put away long ago if you have ac-heating. Gone is the grimy, murky, & noisy sound - now they project into the room with vigour and vivacity.

I briefly considered a complementary current source to isolate both filament terminals, but soon realised a much better circuit with current sinking on the negative, and a gyrator (aka electronic choke) to reject ripple on the positive supply. The schematic tells the tale.



The first (mains) trafo has its output centre tap grounded (300B ground) to eliminate dc and common-mode mains noise. It also serves to keep diode noise and current pulses from feeding back up the mains & into other equipment. Thanks (again) are due to Lynn Olson for voicing this idea.

The second (toroid) trafo is fullwave schottky rectified to minimise the number of clanking diodes compared to a bridge. The target output is 8,7 to 9,5V with 1V of ripple. The high ripple voltage is mandatory to increase the conduction angle and reduce peak current. Ripple is cut to the mV level by the action of the gyrator Q1, T1 and associated passives. The NPN-PNP complementary feedback pair (CFP) works somewhat like a darlington here but with the big difference that when faced with out-of-band signals (limit typically set by C13) the CFP looks like a high (=R3, 1,6K) impedance. So one can try different values for different amplifiers. C13 is also required for CFP stability since there is no output capacitor for the CFP (caps between the filament terminals or to the filament supply return will destroy the performance of this circuit BTW). Be sure you have a scope at hand if you mess with the CFP parts values/types. And for T1 try to get some Zetex FZT1051As from Farnell or Digikey - substitutes will degrade performance if the high Hfe over wide Ic is not available, together with low Cob. For Q1, choose a TO-220 part with Cob < 75pF and Ft > 30MHz. For Q1 and T2, use a 5 degC/W heatsink or better - and mount the sink in free air.

The filament return side is current regulated by T2, T3 and the sense resistor. This simple circuit answers perfectly the requirement for high audiofrequency impedance. Simply select the current sense resistor R4 to give 5,2V across your 300B filaments. 0,5 ohm is a starting value for 1,1A to 1,3A 300Bs, and you then add parallel values to home in. The voltage drops about 5% when warmed up, due to the tempco (-2mV/deg C for Vbe) of T3 - so mount this part in free air.

Do not use LM317 in this position if you want good performance. The dropout and ac impedance are markedly worse, it is noisy, and its thermal cutout preserves a tenpenny chip at the expense of your precious triode.

when set up correctly, there should be about 2V across each power transistor for a 9V supply. Don't let it get below about 1,6V for best ac-performance, and at the other extreme, take care not to use a raw dc supply much above 9V or you will roast the underside of your amp, and probably overstress the transistors. And If you are not comfortable about setting currents in your expensive triodes, or are unfamiliar with stability testing of transistors, or if you cannot confidently calculate and test the thermal design/heatsinking of power transistors, please don't attempt to build this circuit. It will be cheaper in the long run to buy Guido's module and avoid wrecking your DHTs.

Hi

This is an impressive setup

What is curiious, is that you use current sources, and finally "short" the audio current to ground using a cap

Now very likely the filament supply is less in the signal, that function is taken over by the caps

it shows that traditional regulators are not optimised for this task, and that a decently designed supply meets the requirements way better.

Personally I am in favour of high impedance across the filaments, but as you can see, many ways lead to rome....(they are all long though :-)

I guess balancing the sources is tricky, not ?

Oh, one last word, the Sovtek is not the best sounding tube, to be honest I iuse these for first time testing only......

cheers

[Pjotr]

Originally posted by Guido Tent

What is curiious, is that you use current sources, and finally "short" the audio current to ground using a cap

Now very likely the filament supply is less in the signal, that function is taken over by the caps

Only one side of the filament is shorted to ground with a cap. Does not look that curious for a self-biased set up. Apart from the current sink the other side is loaded with resistors, which are relatively high compared to the filament resistance.

Cheers

[Guido Tent]

Originally posted by Pjotr


Only one side of the filament is shorted to ground with a cap. Does not look that curious for a self-biased set up. Apart from the current sink the other side is loaded with resistors, which are relatively high compared to the filament resistance.

Cheers

Hi

Yes. The resistors look like to be the bias resistors. I am not so fond of cathode bias, by the way. I'd rather go for fixed bias and leave the cap out. That is still possible with the proposed filament supply.

cheers

[fdegrove]

Hi,

Guido,

Regarding the preference (subjective thing, I know) , when the + side of the heater is grounded won't that change the bias with the respective heater voltage amount?

I mean, +5V side grounded means -5V added to negative bias or do I see this the wrong way?
If I am correct that may well account for the audible difference I suppose...

Cheers,