"Another Poor Man's" Fairchild 660/670

[lewilson]

I love your post! I am learning a bunch.  Many thanks..
    Well in my case my transformer is about ten to one  with 60k on the primary. My anode voltage is 220. Mine is running about 6ma per tube. 25 ma. per side.
  But regardless, the plate resistance of the tubes is halfed when you put them in parallel. So  in total you have 4 tubes in parallel for very low impedance into the 10k . Also capacitance goes up. Maybe there is the calculation difference?

I admit I just skimmed your last post. I must get to work . ill get back home with my calculator and read it closer.

 

[MeToo2]

I love your post! I am learning a bunch.  Many thanks..
   Well in my case my transformer is about ten to one  with 60k on the primary. My anode voltage is 220. Mine is running about 6ma per tube. 25 ma. per side.
  But regardless, the plate resistance of the tubes is halfed when you put them in parallel. So  in total you have 4 tubes in parallel for very low impedance into the 10k . Also capacitance goes up. Maybe there is the calculation difference?

I admit I just skimmed your last post. I must get to work . ill get back home with my calculator and read it closer.

I openly admit to learning too. So be careful who you learn from The 4 tubes in parallel is already taken into account [I hope] Remember that the amp is running push pull for large drive voltages effectively powering only one of the secondaries at a time. So when driven hard, only one half of the windings are in effect at a time. That also means the peak power and mean power per tube is halved as only one half of the amp is running at a time.

So a 10K to 600 transformer has an impedance ratio of 16.6667:1 = turns ratio of 4.08:1. But if only one half is conducting the reflected impedance back into the signal amp is (4.08/2)^2 = 4.16*600 ohms = 2500 ohms, spread over 4 tubes = equivalent of a 10K load on a single tube.

Your 60K primary on the transformer makes all the difference for a 6ba6 compared to the 10K:600 transformer shown on page 1. 60K:600 = impedance ratio of 100:1 = turns ratio for a full transformer of 10:1 = half transformer turns ratio of 5:1 = Rload of 600* 5^2 * 4 tubes = 60K load impedance. So basically your build is already approximately what I was proposing as a mod......

Also the grid stopper and screen grid stopper are "per tube" connected as close to the socket as possible. So yes, although the total system capacitance goes up, the RC filter for each tube remains the same.

 

[lolo-m]

I openly admit to learning too. So be careful who you learn from The 4 tubes in parallel is already taken into account [I hope] Remember that the amp is running push pull for large drive voltages effectively powering only one of the secondaries at a time. So when driven hard, only one half of the windings are in effect at a time. That also means the peak power and mean power per tube is halved as only one half of the amp is running at a time.

So a 10K to 600 transformer has an impedance ratio of 16.6667:1 = turns ratio of 4.08:1. But if only one half is conducting the reflected impedance back into the signal amp is (4.08/2)^2 = 4.16*600 ohms = 2500 ohms, spread over 4 tubes = equivalent of a 10K load on a single tube.

Your 60K primary on the transformer makes all the difference for a 6ba6 compared to the 10K:600 transformer shown on page 1. 60K:600 = impedance ratio of 100:1 = turns ratio for a full transformer of 10:1 = half transformer turns ratio of 5:1 = Rload of 600* 5^2 * 4 tubes = 60K load impedance. So basically your build is already approximately what I was proposing as a mod......

Also the grid stopper and screen grid stopper are "per tube" connected as close to the socket as possible. So yes, although the total system capacitance goes up, the RC filter for each tube remains the same.

Be carefull, I think that you're not completely right on one particular point :
The 660/670 varimu part is in push pull., right. But it's not push then pull if I understand correctly what you wrote, it is push and pull at the same time. That's why poeple used to say Class A push/pull.
Even the SC amp works like this because of the feedback but more assymetricaly...

Safe triode mode for 6BA6 is : pin2 tied to pin7 and pin5 tied to pin6. That's what I did but honestly, most of the folks here don't bother with that...

 

[lewilson]

Yes, this is exactly how mine is.

 

[MeToo2]

I openly admit to learning too. So be careful who you learn from The 4 tubes in parallel is already taken into account [I hope] Remember that the amp is running push pull for large drive voltages effectively powering only one of the secondaries at a time. So when driven hard, only one half of the windings are in effect at a time. That also means the peak power and mean power per tube is halved as only one half of the amp is running at a time.

So a 10K to 600 transformer has an impedance ratio of 16.6667:1 = turns ratio of 4.08:1. But if only one half is conducting the reflected impedance back into the signal amp is (4.08/2)^2 = 4.16*600 ohms = 2500 ohms, spread over 4 tubes = equivalent of a 10K load on a single tube.

Your 60K primary on the transformer makes all the difference for a 6ba6 compared to the 10K:600 transformer shown on page 1. 60K:600 = impedance ratio of 100:1 = turns ratio for a full transformer of 10:1 = half transformer turns ratio of 5:1 = Rload of 600* 5^2 * 4 tubes = 60K load impedance. So basically your build is already approximately what I was proposing as a mod......

Also the grid stopper and screen grid stopper are "per tube" connected as close to the socket as possible. So yes, although the total system capacitance goes up, the RC filter for each tube remains the same.

Be carefull, I think that you're not completely right on one particular point :
The 660/670 varimu part is in push pull., right. But it's not push then pull if I understand correctly what you wrote, it is push and pull at the same time. That's why poeple used to say Class A push/pull.
Even the SC amp works like this because of the feedback but more assymetricaly...

Safe triode mode for 6BA6 is : pin2 tied to pin7 and pin5 tied to pin6. That's what I did but honestly, most of the folks here don't bother with that...

OK. Understand your comment now about safety on a 6ba6. Yes I would also recommend connecting 2&7 on a 6ba6.
On a 6k4p-ev they're apparently already connected inside the tube.

Class A? really? what makes you say that?

For small signals I'm positive it's class A, when not being driven too hard.

But for larger signals (up to +27dBm) doesn't it depart into Class B? = class AB overall.

The signal amp output transformer is 1+1:9+9. The supply is 240V regulated. The spec says that it clips at +27dBm = 25.42V peak. Which for 9:1 ~ 240V or the full supply voltage. The other half then has to be pretty much off doesn't it? Otherwise it wouldn't clip?

What's the bias point then and quiescent current?

Imagine you pump Dutch techno square waves through it at +10dBM  :
When the control voltage goes down to -38V for 10dB compression are they still running class A to full output?

 

[lolo-m]

Of course you can make the varimu part going into class B with some extreme setting (I mean no SC signal = threshold all way down or so). But the sound is really distorted like this and completely unusable... except in some case but I haven't found yet a really usable sound with this.
You are in a looped system. Except during the first 0,2ms (or 0,1ms) where cliping can occur (and will occur in extreme situations) the Grid 1 bias (CV is negative DC) will collapse to a point where the tube is in clean push-pull mode. The 240V are possible without any effort as you have to measure this between plates and not between the center tap of the transformer and its tips... BTW in the case of a push-pull stage, there's another point you must take into account : the transformer isn't used there as a voltage converter but as a current to voltage converter.
I remember my surprise the first time I took measures on my varimu's transformers seeing how low were the voltage drop through the primaries. It was a few volts (or mV) closer to U=Rtr X I where Rtr is resistance of windings than U= RrefLoad X I where RrefLoad is the refflected load resistance... So there is nearly the total PSU voltage at any time on the plates.

 

[MeToo2]

Of course you can make the varimu part going into class B with some extreme setting (I mean no SC signal = threshold all way down or so). But the sound is really distorted like this and completely unusable... except in some case but I haven't found yet a really usable sound with this.

Sure. But the extremes are where the tube melts, which is precisely what I was trying to look at.
I wasn't looking at the small signal model at all.

You are in a looped system. Except during the first 0,2ms (or 0,1ms) where cliping can occur (and will occur in extreme situations) the Grid 1 bias (CV is negative DC) will collapse to a point where the tube is in clean push-pull mode. The 240V are possible without any effort as you have to measure this between plates and not between the center tap of the transformer and its tips... BTW in the case of a push-pull stage, there's another point you must take into account : the transformer isn't used there as a voltage converter but as a current to voltage converter.
I remember my surprise the first time I took measures on my varimu's transformers seeing how low were the voltage drop through the primaries. It was a few volts (or mV) closer to U=Rtr X I where Rtr is resistance of windings than U= RrefLoad X I where RrefLoad is the refflected load resistance... So there is nearly the total PSU voltage at any time on the plates.

Again I can understand the small signal model and the overall feedback looping. The discussion I documented was about a large signal model. Under normal operations the discussion on power and large signals is irrelevant. Everything is pretty much turned off. The compression will indeed back off the whole signal so that the voltage swings reduce. That's the point of compression.

All I was trying to look at was the safety of the (abnormal) situation where the compression is off (for example the input pot of the control amp is turned to 0 or the DC threshold turned way up exactly as you suggested) and the input is blasted, so the signal amp does go into saturation or when someone overloads the load with less than 600 ohms. I was concerned about the pentode tubes fusing and think I know how to prevent that. I don't really care how it sounds at that point (maybe "pop" :-\) . The screen grid stoppers should have virtually no impact during normal operation. The plate current  is 2* the screen grid current on a 6k4p during normal pentode operations (Vplate >20V). Thus the current through the load is 3* screen grid current. If the screen grid stopper is even as high as 10K then that is effectively in series with 40K*3 = 10K is series with 120K < 1/13*Vload = pretty negligible voltage drop of Vscreen_grid compared to Vplate. The screen grid stopper resistor only comes into play when Vplate <20V and the plate current falls off at saturation, and the screen grid current rises dramatically. Practise will prove the theory or not and whether it has an (undesirable) linearising effect on the voltage gain.

Not sure I understand your comments on a "current to voltage converter." A transformer transforms power. Power in is approx power out (minus imperfections). It reflects the load back in to the input as the square of the turns ratio * the output load. A current to voltage converter is called a resistor in my book. In the small signal model, the variable mu effect comes about because of the variance of Mu (voltage gain) over the large operating range of Vgrid, not the variance of Gm (transconductance). If there's virtually no current (low signal) then of course the voltage will remain at near PSU voltage. Most of the 240V supply is dropped across the valve (the valve is nearly switched off) and very little is dropped across the transformer primaries. That is exactly why I think it is running class AB. But then again you seem to have a different opinion. OK. I respect that. I'll find out when I build it. I'll be interested to measure the difference between DC and AC current and voltage (because of course a transformer secondary has virtually no resistance to DC)

Maybe have a look at this http://www.freewebs.com/valvewizard1/pp.html which attempts to explain the transition in impedance from class A to B operation.

 

[lolo-m]

Not sure I understand your comments on a "current to voltage converter." A transformer transforms power. Power in is approx power out (minus imperfections). It reflects the load back in to the input as the square of the turns ratio * the output load. A current to voltage converter is called a resistor in my book. In the small signal model, the variable mu effect comes about because of the variance of Mu (voltage gain) over the large operating range of Vgrid, not the variance of Gm (transconductance). If there's virtually no current (low signal) then of course the voltage will remain at near PSU voltage. Most of the 240V supply is dropped across the valve (the valve is nearly switched off) and very little is dropped across the transformer primaries. That is exactly why I think it is running class AB. But then again you seem to have a different opinion. OK. I respect that. I'll find out when I build it. I'll be interested to measure the difference between DC and AC current and voltage (because of course a transformer secondary has virtually no resistance to DC)

Well, quick measures on my PM670. Nothing plugged on output. No compression. So max current on the 15K (approx) reflected load (SCamp input). Voltage drop : less than 1V. Accoding to theory, current should be less than 0,067mA and it is a lot more (a few mA).

Current in primary creates a magnetic field in the transformer. This magnetic field creates a current (or a voltage if there's a load) in the secondary. More than a power converter, a transformer is a current converter. If there's a load it becomes a power converter... Build quickly a push-pull stage and you'll understand  .
Another point is that you theory doesn't work if there's no load at all. Infinite load on secondary would mean infinite load on primary so no current at all on the plate and so nothing happening. But there's the normal polarisation current on the plates and plate current follows grid 1 voltage as it should.
Maybe are you confused by the PRR's varimu where the plates load are actual resistors. I confess it did confuse me when I started to understand varimu's  ;D.

 

[MeToo2]

Not sure I understand your comments on a "current to voltage converter." A transformer transforms power. Power in is approx power out (minus imperfections). It reflects the load back in to the input as the square of the turns ratio * the output load. A current to voltage converter is called a resistor in my book. In the small signal model, the variable mu effect comes about because of the variance of Mu (voltage gain) over the large operating range of Vgrid, not the variance of Gm (transconductance). If there's virtually no current (low signal) then of course the voltage will remain at near PSU voltage. Most of the 240V supply is dropped across the valve (the valve is nearly switched off) and very little is dropped across the transformer primaries. That is exactly why I think it is running class AB. But then again you seem to have a different opinion. OK. I respect that. I'll find out when I build it. I'll be interested to measure the difference between DC and AC current and voltage (because of course a transformer secondary has virtually no resistance to DC)

Well, quick measures on my PM670. Nothing plugged on output. No compression. So max current on the 15K (approx) reflected load (SCamp input). Voltage drop : less than 1V. Accoding to theory, current should be less than 0,067mA and it is a lot more (a few mA).

Current in primary creates a magnetic field in the transformer. This magnetic field creates a current (or a voltage if there's a load) in the secondary. More than a power converter, a transformer is a current converter. If there's a load it becomes a power converter... Build quickly a push-pull stage and you'll understand  .
Another point is that you theory doesn't work if there's no load at all. Infinite load on secondary would mean infinite load on primary so no current at all on the plate and so nothing happening. But there's the normal polarisation current on the plates and plate current follows grid 1 voltage as it should.
Maybe are you confused by the PRR's varimu where the plates load are actual resistors. I confess it did confuse me when I started to understand varimu's  ;D.

Ok to be really exact: a transformer is really a flux convertor between two mutually linked inductors, and magnetic flux is derived from I squared. For a resistive load that gives I squared R = power. None of the diagrams deal with driving inductive loads, like a loudspeaker, as this is a line amp. Infinite load resistance in a transformer while powered up can lead to dangerous voltages on the secondary as there's nothing to control the flux in the primary. I wouldn't advise anyone reading this trying that test on for example a step-up (input transformer). You may kill yourself.

I've never suggested the amp is running pure class B. There'll always be a bias current in class AB (even if it is tiny) which leads to your correct tracking. But it might still depart into class B at the top of the cycle.

The DC resistance of the output transformer will always be near zero, so for DC conditions the plates will sit at or near Vsupply (= DC resistance * bias current) just as you have measured.

As for the impedance load question for AC, it really depends on the biasing point and the applied load which answer is correct: class AB or A. It's as simple as that. When running in class B (even for just part of the waveform) then only half the windings are working. In class A then both sides are working. The impedance that the valve sees can even change partway through the height of the waveform (as the other partner shuts off).

Don't worry, I'll be very conservative with resistances and voltages when I first build this to start with to see how it is really running. And I do intend to publish any curves I manage to plot for both large and small signal behaviour [even if I'm completely wrong. That's not important.].

 

[MeToo2]

My batch of 6k4p-ev have arrived  ;D

I've now plotted the plate and transfer characteristics for triode-mode connection 6k4p-ev, as none of the data sheets I could find dealt specifically with triode mode and especially large negative grid voltages, which is what is interesting in this project.

I must say the curves look very similar in shape to published data for a 6386 (except of course the max power and max anode voltage are different.) Since I believe a large proportion of the character from "variable mu" compression comes from the shape of the AGC curve (and thus the variability of mu) I think a pair of matched 6k4p-ev (at around 1-2 euros each) are potentially a very good option to substitute for a 6386 in an "affordable" 660 circuit. You may disagree.

I'll try to do two things: publish the raw the 6k4p data, and plot the data on the original 6386 curves.
Anyone wanting the raw spreadsheet data just PM me.

Up until now (for DC behavior) I see no reason for the screen grid stoppers, provided you keep the major voltages and current within limits.

Note there's a few caveats....
Sample of 1 valve measured. No idea about manufacturing tolerances.
Curves were measured manually at DC. No AC characteristics measured (although it is a valve for VHF)
There's some power supply dip clearly visible at low grid voltages, which kinks the curves near Vg=0V.
I kept max power (DC) to around 3.6W. Obviously for transients you could easily exceed these values at higher plate currents, but that's not an issue as long as the mean power remains within range. Cathode current is recommended at 20mA max.
The curves are good enough for me. If anyone has a proper curve tracer then fair do's and please publish.

Here's the 6k4p-ev curves I measured:

http://i958.photobucket.com/albums/ae69/MeToo2_Prodigy/6k4p_triode_mode_data_transfer.jpg
http://i958.photobucket.com/albums/ae69/MeToo2_Prodigy/6k4p_triode_mode_data_plate.jpg




Here's the data plotted on the same axis as the 6386 curves on the data sheet.
http://i958.photobucket.com/albums/ae69/MeToo2_Prodigy/compare6k4p-to-6386-plate.jpg
http://i958.photobucket.com/albums/ae69/MeToo2_Prodigy/compare6k4p-to-6386-transfer.jpg

 

[schlamn]

Good work sir... Thanks for sharing your progress/data here 

 

[syn]

Thanks a lot, I'm watching this with great interest as I'll be building APMF660 (dual)  soon-ish.
Best

 

[jackies]

I was wondering, rotheu, the original topic starter, listed his location as "CCCP", is he really from the former USSR? And if he is, why wouldn't he try some russian tubes, which should be plentiful anywhere within the former russian empire..
:

 

[MeToo2]

Just FYI. I haven't gone to the lengths of JJ Electronics matching the 2 halves of a tube at 5 points along the curve. Just checking hteir basic self-biassing at the correct level.

Out of a batch of 30 tubes I managed to get 2 sets of 8 tubes that seem to match approx ±1% so I can build a dual mono unit as planned. There were a couple of out-lyers (low or high gain.) If that experience is repeatable you should probably be thinking on buying approx twice the number of tubes that you'll need (and try to get them all from one batch)

 

[syn]

I have 6k4p-ev x 30 and 12 x 6n5p. I just have to finish a couple of things that i'm doing at the moment, and I'll be starting the build. I plan to use russian tubes throughout the unit apart from
12au7. This (dual) unit is going to suck a lot of  A.

 

[syn]

6kp4=0,325x8=2,6
12au7=0,3x1=0,3
6n2p=0,365x1=0,365
6N6P=0,82x1=0,82
------------------------------
total=4,085A

these are per ch. 6.3V heater.So 4.5A per a ch. would do for the heaters I guess.plan to use lm338k per ch.
probably two "independent" PTs and PSUs.
I still have to calc. for the plates supply.

 

[Kingston]

question about these pentodes in triode modes.

Seems to me that 6BA6 is not exactly the equivalent to the 6K4P as has been mentioned often.

for example in the PM670 thread Bluebird says he connected "Pins 6,5,2 all together for the plate" on the 6BA6 for the triode mode. That is plate and both of the top grids on the data sheet connected together and seems to work fine.

For the 6K4P this can't be done. grid 3 hardwired to cathode. I should then just connect grid 2 to plate, and that's a triode (as pictured above as well).

But what's the effect of this compared to the 6BA6 as triode where no grid is connected to cathode?

 

[MeToo2]

6kp4=0,325x8=2,6
12au7=0,3x1=0,3
6n2p=0,365x1=0,365
6N6P=0,82x1=0,82
------------------------------
total=4,085A

these are per ch. 6.3V heater.So 4.5A per a ch. would do for the heaters I guess.plan to use lm338k per ch.
probably two "independent" PTs and PSUs.
I still have to calc. for the plates supply.

Yes I reckoned on around 4 amps per channel just for the heaters. Although you could run some at 12.6V (e.g. 12au7) and the rest at 6.3V. The power supply is still a question mark for me. I'm actually thinking about 4 separate supplies [heater, signal amp HT, Control amp HT1, control amp HT2]. Note that the initial heater current at switch on can be significantly higher than during normal operation (0.5A or so per tube for several seconds) so if you are going to regulate the heater current you might want to include an over-current limit, although I read that the LM338 had that built in. Depending on how hot you decide to run the 6k4p's the XPWR039 or maybe the XPWR063 from Edcor may be worth a look to at least power the signal amp section & associated heaters. Then a separate tube amp transformer for control amp HT1 & control amp HT2 + associated 12V heaters [ala Gyraf G9].  I was not planning on regulating or rectifying the heater power at all if I went for the XPWR039: just raw 6.3V AC. Risk of hum is obvious, although the original had no conditioning on the heaters. If you have better suggestions with the XPWR063 + TL783/LM338 regulation please let me know, as I've been puzzling for some time on this one.

question about these pentodes in triode modes.

Seems to me that 6BA6 is not exactly the equivalent to the 6K4P as has been mentioned often.

for example in the PM670 thread Bluebird says he connected "Pins 6,5,2 all together for the plate" on the 6BA6 for the triode mode. That is plate and both of the top grids on the data sheet connected together and seems to work fine.

For the 6K4P this can't be done. grid 3 hardwired to cathode. I should then just connect grid 2 to plate, and that's a triode (as pictured above as well).

But what's the effect of this compared to the 6BA6 as triode where no grid is connected to cathode?

The "traditional" way of wiring a pentode as a triode is to connect grid3 to cathode (some valves already have this connection internally) & to connect grid 2 to anode (plate). There are others who say that grid3 should be connected to anode, or even used as the main grid..... I don't subscribe to that view, but there are others out there who do. see e.g. http://www.webace.com.au/~electron/tubes/triode.htm As he notes there seems to be very little manufacturer's data out there, which is why I produced and published my own curves.


As an aside.... I'm asking Edcor about an XSM50K/600 and a XSM600/50K. Please note that this is experimental, but I believe there's some value in increasing the output impedance load on the signal amp to mean that it runs longer in class A operation even during significant compression. Its gonna cost me $40 per transformer in one off set up charges but it should then be available for everyone as a standard product. We'll see if this idea works out in practice, but it's gonna take a week or two before I can report back......

 

[syn]

I think I'll go regulated 6.3V all the way through."Rule" that caps needed are in uF about 10x what we have in mA  bring us to ~45000uF per channel for the heaters only.
I linked to my edits of  the possible schematics to be used for the psu.
They should be considered as a concepts only, please, don't pay attention to the values shown as nothing is calculated, yet. This is just to help (me) us decide what route is to be chosen PSU wise.
Schematics are originally Gyraf G9 and FotisGR slow blow. Big & expensive choke from the slow blow could be substituted with a resistor, to a more or less similar performance (wise man said).Also i think, slow blow HT PSU could be tapped for 240V and 300V by addition of 1r+1c. It is not shown on the schemo but you know what I mean. The last schemo  was pointed to me by Rotation I  haven't build anything with it yet, but it looks good. It works good up to 300V so 240V should be no problem. Again, with two tip50 and surrounding circuit adjustment we could get  300 & 240V from single secondary of the PT.
If  I understood Rotheu correctly for 6kp4 build 100V is not needed (think 240V goes to 6kp4 plates as well).
I'll go with 10k:600 Carnhills, and a custom wound PT.
Thanks

 

[MeToo2]

OK. Thanks. keep us updated on your thoughts. I'll do the same.

My concern about tapping a single HT supply and regulating the heaters is the amount of power you may have to dump. Also running the 6k4p @ 240V might be very hot in the push pull signal amp (I'm speaking hypothetically here, as I haven't tested yet. But the push pull output transformer can go into "auto-transforming", even if little or no current is passing. This means that the fly back voltage on the plate of the tube that is turning off can easily double compared to the static plate voltage. Max plate / anode voltage for a 6k4p is 300V. You could easily end up with 480V peak AC there for 240V DC feed. I haven't tested that at all yet to see what really happens when built and it is running in class A, so I could be completely wrong but I think it is a good starting point to design push pull to operate at 1/2 Vplate max. However the original 660 seems to ignore this rule too for a 6386)

Although you can get the full power/current out of single 240V tap on a standard tube transformer e.g. 200mA@240V AC = 380V peak, which is then regulated, you may easily have to dump 380-240*4mA*16 ~ 10W plus for two channels of signal amp. Same for the heaters. Say 3V drop across the regulators @ 5Amps. That's another 15Watts of heat. Plus the heaters themselves 6.3V @5A = 30W. Could get awful hot.

custom wound PT.

Custom wound pt could make sense if a few of us were building it. Then we could have separate and optimized HT taps for push pull control amp, control amp, signal amp, and a decent thumping heater supply. The HT regulators cost very little themselves. Pennies each for a tl783. So multiple G9 psu route seems feasible. But where ever I look I can't find exactly what transformer I want. Thing is, at the moment I'm not exactly sure what I want as I feel a need to tinker with the voltages. hmmm. Edcor seem incredibly accommodating though and only charge $40 one time set up. But maybe I just need to buy one of these toroids with lots of taps e.g. http://www.analogmetric.com/goods.php?id=1372 plus a separate heater transformer, or maybe better to get two of these T30 60W http://www.analogmetric.com/goods.php?id=78 to start with. They are sold on eBay too. Looking at this further, 2 off T30 might just work. 0-6.3V(1.2A) x2 covers the heaters for 8 off 6k4p's exactly = 1 channel of signal amp. The 0-15V(1A) could be rectified & regulated to 12.6V to supply the heaters for one channel of the control amp (by my calculation 0.7A @12.6V). 100mA should be enough for one channel of signal amp (~30-50 mA class A) leaving ~50mA for one channel of the control amp for 240V regulated (12ax7 12au7) & 300V (ecc99) for a total > 10W, which should do at a stretch if using ECC99's in the control amp (5W max power each anyway).