Build Thread - Modern Power EL84 P-P amp

[linuxslate]

I have decided to take the concepts from my Sub-miniature Tube amp project, and apply them to a larger build.

This includes using all solid-state, switch mode power supplies for all of the needed voltages. Please see this thread for more about that.

This thread will document (and request comments/help) for this project.

Overall, I am building a very basic 6BQ5/EL84 Push-Pull design. Perhaps best thought of as a copy of a Packard-Bell console amp chassis. It could also be thought of as the same design of the Sub-miniature Tube amp build around miniature tubes.

A schematic will follow. When I post the schematic, I'm sure I'll get lots of comments suggestions (and I do have some questions), but we'll leave that until I get a little closer to actually soldering.

For now, just as a teaser, and to show that I am getting serious, here is a shot of some of the parts I have collected so far:



First, of course I have to mention what is not pictured:
-- Obviously the Chassis. That is on it's way.
-- The chassis is supposed to include an IEC chassis-side connector with integral fuse holder.
-- Resistors. I have lots of resistors, and I'll pick needed values at build time.
-- Capacitors other than the (shown ) main coupling caps. Again I'll pick needed values at build time.
-- I am currently planning on having a built-in USB DAC. That is also on it's way.
-- The amp is going to have another feature that I am not revealing at the moment. It depends on (yet even more) parts from China, and my evaluation of them when and if they arrive.
-- Misc common parts such as wire, grommets, tie-wraps, heat shrink, screws, etc.

I should also mention that due to the use of the power supply modules, there really aren't a lot of other parts.

Other comments:
-- Yes, I pained the OPT's. They are new, but not cosmetically perfect, so I am attempting to make them look cool. The shields were removed, painted, and black glitter was added.
-- The small buck converter with the green heat sink (left center) is the filament power supply. I may substitute a slightly larger one with a higher current rating (on hand).
-- I am even building my own power cord. The pictured fabric-covered line cord is cut from a 3 outlet power strip. The one shown is actually purple. I have another one in another color, and may even buy another if I see one with a cord I like.
-- Hey! Who left the clip from the bag of sockets in the picture! (That's not part of the amp.)

Next:
The tubes are not the ones I planned to buy. Story (and request for suggestions) to follow.

 

[linuxslate]

Description of the power supplies...

Down the right side:
100-240 V AC input, 12V DC, 8.5A output power supply. This provides the main 12V power, and Isolates the entire amp from the AC mains.
(Module with a yellow transformer) -- Adjustable HV supply -- Plate Voltage for the EL84's.
(Large module with a greenish toroid in the middle) -- Screen and Driver section supply (~280V) This also includes a low voltage supply intended for the filaments, but due to ripple when heavily loaded, the low voltage section will be set to 5V, and used for the DAC, and other digital accessories.

Down the left side:
(Three small toroids in the center) -- Positive/Negative module -- Bias Supply.
(Green heatsink in the middle) -- Filament Supply. May use a different one.

All are adjustable and regulated. There will be little to no change in the amp's operation over a range of 100 - 240 V AC input.
While I am not including a 12V input, the amp could also run from any source of 12V DC that can supply ~ 8A.

 

[Lavane]

 

[linuxslate]

Here's a tentative layout:



I have much more detail and measurements in the actual drawing, but those are suppressed for this quick picture.

The plate supply and the bias supply are above the chassis, and inside a removable cage. The cage will be made out of the same mesh as I used on the Sub-Mini Tube Amp (It is from a metal mesh garbage can). This time, I am going to attempt to weld threaded rod in the corners of the cage instead of epoxying. The rods will provide extra material for welding, while forming the corners of the cage. If the mesh vaporizes when I try to weld it, I'll either have someone tig weld it, or resort to something else. Perhaps Lexan would look better anyway.

There will be holes to allow plate voltage and bias adjustment from the top.

There are no cathode resistors; I will use the socket adapters / bias meter gadget that I have. No covers need to be removed to bias this way, and there is no high voltage involved.

Of course, I'll still need access to measure the plate voltage to adjust the HV supply.

All other power supplies are under the chassis. This means the unit will need to be opened to adjust the filament voltage or screen voltage.




The rear panel is much more rough at this point. I'll know more when I have the remaining parts in hand.

The RCA jacks will be as follows:

DAC OUT DC Coupled Input Capacitor Coupled Input


To use the built-in DAC, RCA jumpers will couple the first 2 sets of RCA jacks. The unit could be used as just a USB DAC (in a really heavy and oversized box) without power. (Depending on power supply jumper on the DAC board.)


No front panel drawings at this time, because it depends on the other items I have on order. There is only an illuminated power switch on the front as of right now.

 

[linuxslate]

About the tubes...

As you can see from the pictures in the 1st post, I ended up with JJ EL84's.

I had 2 intentions, and it turned out that they were incompatible.

1.) I wanted to support a local retailer - (Minor plug: http://www.thetubecenter.com/). He sells at local hamfests, and I could buy without the cost or delays of shipping.
2.) I had planned to use Russian "surplus" tubes. (New, but perhaps older build). 6P14P, Probably branded as Sovtek, Reflektor, Svetlana. I figured there is more supply than will probably ever be used, they are pretty much indestructible (incase of a wiring error, bias problem, etc.). In a small lot, I could get them for well under $10.00 a piece.

However the guy from thetubecenter tells me he does not touch the Russian stuff. I stuck with supporting the local guy, and paid ~$14 a piece for the JJ's.

The only recent build 12AX7's he had were completely unmarked Chinese. He says they are excellent for "low noise". (I think that's kind of like "Nice personality" meaning "unattractive"). I'm worried they are probably slightly more microphonic than a condenser mic element. The mica is jammed up against the side and bent in one of them. I even think the entire guts are a little crooked.

I'm sure they will be fine for getting it up and running, and I can always order some Russian stuff later.

 

[linuxslate]

Time to post a preliminary schematic:

Important Notes:

1. The OPT's I have only have an 8 ohm sec tap. The 4/8 taps on the drawing are due to cut and paste. (Physical drawings shown above correctly show only one set of output posts.}
2. I may get rid of the 1 ohm resistors on the 6BQ5 cathodes. I plan to use socket adapters to set/check bias. Another thought was to replace them with fuses.
3. I'm not sure the 1K resistors on the grids are needed either. It's kind of interesting that I basically end up with no passive components around the 6BQ5's.
4. Input coupling caps will change to 0.1uF based on availability. Shouldn't matter, as those are outside the feedback loop and are intended to only block DC should an input that is not know to be zero referenced be connected.
5. EDIT: Only the ground on the left of the Mains supply is actually an earth ground (Mains ground). All others should be drawn as chassis ground. The 12V PS isolates the input and output, including the grounds.

Thoughts/comments appreciated.

 

[PE9ZZ]

Nice. Make sure the SMPS's are very well filtered. If switching products end up in the power lines beating against each other you will have SMPS noise in your sound which will be very ugly. Also, where's the volume control?

Tjerk, 9ZZ

 

[tonyg88]

Very cool! Would this still be considered a CHIFI build?

 

[linuxslate]

where's the volume control?

No volume/tone/input select switches. This is intended to be driven from a digial source (be that via the internal HiFi DAC, or line audio from the source device). The volume control will be in the digial source device. Otherwise you end up with 10 volume controls in series. (perhaps slight exaggeration, but it is at least 3).

Same story for tone control. Any equalization is done at the digital source (DSP) -- i.e. PulseEffects for us Linux users.

The RCA jumper avoids input switches.

The only control will be the power switch. (Not shown in schematic. -- There is also a fuse in the IEC socket, also not shown.)

In addition to a fuse in the IEC socket, the Line-to-12VDC amp is fusted at the input, and the HV (Plate) PS is internally fused.

 

[linuxslate]

So the discussion of SMPS/Buck/Boost is in my other "Modern Power" thread.

Trust me -- I've gotten lots of comments about SMPSs.

I was hoping for comments about the rest of the design here.

Assume (right or wrong) clean DC at the set voltages.

Comments on the voltages themselves (for example screen voltage vs plate voltage) are certainly welcome.

While I do hope to be maximizing the power that can be safely gotten out of 6BQ5s (by hard-grounding the cathodes), I am not going for horsepower. More for efficiency.

I don't intend to run the plates any where near datasheet maximum.

 

[gadget73]

Ballpark without doing any detailed analysis, the output section looks OK to me. Having the screen supply ~40v lower than the plate should make the tubes a lot less inclined to arc and be otherwise stupid. Vaguely possible you'll need more than -15v of bias but it certainly wouldn't be more than -20v. I don't notice an adjustment for bias or balance though.

Not entirely sure the resistors on the phase inverter and voltage amp are right. Just eyeball it looks like the plate resistor on the first section is too large, and the two on the inverter may be too large as well. Thats just a thought based on other designs. See maybe a Fisher SA-100 for some target values. Those ran the same inverter and voltage amp design. I've never been much good at whiteboard designs, I tend to look at other successful things and work from that as a model.

 

[linuxslate]

The bias adjustment is just the voltage of the (-) bias supply. It's been a while now since I had it on the bench, but as far as I remember it would go somewhat under -15V with a +12V input. I just can't remember how much. It won't do -20V.

I'm also not sure if you twist the pot all of the way to the end that the negative buck/boost may do more of the"bad stuff" everybody is talking about. Obviously, any noise there is going directly into the grids and will be amplified big time (although it will be common mode).

I think you said the part about the screen supply wrong. I'll just say that I do not plan a screen voltage so different from the plate voltage as to cause arching or stupidity. I realize the ranges shown on the drawing implicate like 100V possible difference. I wouldn't do that.

There is no balance control.

I realize I'll have to play with the input stage. Especially with it being direct coupled.

I'm relying on sockets that can take some repeated soldering to get that right.

Most of the schematic input and inverter is taken from a Packard Bell console amp schematic.

 

[gadget73]

having the screen lower by 40v or so (320 plate, 280 screen) makes it LESS inclined to be stupid. Its when the plate and screen voltage is very close together that there is more potential for arcing. Still might not be a bad idea to put 100 ohm screen stopper resistors in there.

 

[gadget73]

so I had another thought, you will probably want to elevate the heater supply by some amount to keep heater-cathode voltage differences in the inverter stage under control. I usually aim for about half whatever the cathode voltage is. That keeps everything at a happy medium. Typically a voltage divider off B+ will get it done, resistors can be pretty large in value. It doesn't need to flow any real amount of current. Spitballing it, I'd say maybe a 47k resistor to ground and 220k off the 280v screen supply. That should put you close enough to 50v at the heaters to make everything happy. usually split load inverters run basically in thirds, so 1/3 of the supply voltage across the plate resistor, 1/3 across the tube, and 1/3 across the cathode resistor. With 280v supply thats ~93v at the cathode.

 

[linuxslate]

Still might not be a bad idea to put 100 ohm screen stopper resistors in there.

So based on what you said, I read a little about screen stoppers. Yeah -- Especially with that screen voltage coming from a "hard" (regulated) PS, and not through a "soft" series of voltage divider resistors like in a conventional power supply, I may want some resistance in there especially if/when the amp is run at high power.

...will probably want to elevate the heater supply by some amount to keep heater-cathode voltage differences in the inverter stage...

According to the 12AX7 spec, it's rated for 180V cathode to heater. I get the 1/3, 1/3, 1/3, thing, but that only puts the cathode-to-heater at half the rated max value. (Almost exactly half if you average the 2 ends of the heaters.)

Of course, I was looking at the datasheet for a "real" (GE) 12AX7, not the Chinese ones I got.

I'm also not sure how I would bias the heater voltage up using the buck converters I am using. What you suggested using a 220K/47K voltage divider would bias a seperate winding of a power transformer, but I don't really have one of those in this design. There are ways to do it -- 2 Boost converters 6V apart -- Building my own elevated P/S with something like an LM317, etc, but it is not as simple as biasing a separate PT winding.

I also don't think that the heater voltage was elevated in the Packard Bell console chassis I got the schematic from.

I really like the fact that tubes glow, but in both of these cases, I want the filament to be the only element that does so. I really appreciate the advice, but in this case, I think I'll only adopt the part about the screen resistors. I think I have to take my chances with the cathode-to-heater potential.

 

[gadget73]

Those are DC and signal numbers. Sylvania datasheet calls for a max of 100vdc, and allows up to 180 volts DC + peak. Still not particularly great practice to push that limit if you don't have to.

Often the way old console amps did this was to simply tie the cathode of the output tubes to the heater circuit to get ~15v or so. Not sure how it would need to be done with your heater supply. Normally it will need to reference to B- somehow, but I guess it depends if the buck converter is completely isolated on the output side or not. If it is, then it would be a simple matter to use a voltage divider. If its not isolated you might not be able to do anything with it.

 

[linuxslate]

Another case of not knowing what I don't know. I concider myself a beginner tube amp builder, not a designer (even beginner).

I never really understood why the hum pots in my HK a224 were connected to the cathodes. (I guess I still don't really get how that reduces hum), but I get (now) that they are just trying to get a free ~12-15V of heater elevation.

These buck/boosts do not isolate.

I guess I have some thinking to do. Several options:

1. Add a coupling capacitor between the input stage and inverter to allow running the phase inverter at a lower voltage (thus not needing to elevate heater)
2. Come up with some arrangement of boost/buck in series, taking the 12AX7 heater voltage across the second buck (output to input, not to ground.)
3. Find a small SMPS that does isolate, and has the current for at least the 1 side each of two 12AX7 heaters. Bias with a voltage divider to +280V as you mentioned.
3. Add a sepetate transformer for at least some of the heaters, and again bias with a voltage divider as you mentioned.

If nothing else, I can buy 6V of margin by having a separate 12V buck for the 12AX7's, and just make sure I connect the inverter side to the +12V side. That gets me >10% margin on the Sylvania spec.

Thanks! You made me think, and that is an accomplishment!

 

[linuxslate]

Non-progress update...

1st, I still do not have access to my drill press. I have excuses (legit and non-legit), but posting them will not help anyone.

More significant is the Heater-Cathode voltage issue.

I've been doing a little research:

From a post on el34world.com:

Re: 12AX7's like the tung sol RI that can't handle cathode follower voltage
« Reply #8 on: February 02, 2017, 11:01:52 pm »
Quote from: jjasilli on February 02, 2017, 09:07:03 pm
... However typical 12ax7 (see GE chart) says 100V spec. ...

We're getting into a gray area of "what sheet from what date?"

G.E. 12AX7 says 180v.
G.E. 12AX7A says 200v.
Sylvania 12AX7 says 200v.
Tung Sol 12AX7 says 180v.
Brimar 12AX7 says 250v.
RCA 12AX7A says 200v.

Philips 12AX7S says 100v, but this may be artificially-low to meet claimed low heater-to-cathode leakage current.

The best rule of thumb may be to split the difference between cathode voltage of normal grounded-cathode stages and the cathode voltage of follower stages.

Yes, they do also talk about the DC Heater-cathode voltage, plus program (signal) voltage swing. I get that.

And from a post on music-electronics-forum.com

Many classic amp designs exceed the heater to cathode maximum in one or more preamp stages. It was never a problem for over half a century. Problems with certain Rusky tubes have been noted over the years since. This problem is mostly worked out now and you don't hear about it much anymore with current production tubes. But no promises on that. The typical fix is to just use a Chinese (Shuguang) tube in the socket with the cathode follower.

This is of note because the 12AX7's that I bought are unknown and unmarked Chinese -- "... probably made in the Shuguang factory" (quoting the vendor.)

One other development... I found this mixed in with some Laser Power Supply transformers:



Yeah -- it's pretty wimpy. 300mA is just enough to run the (2) 12AX7's in series (each), and would make it easy to elevate the voltage with a fake CT divided from the +280V. I'm sure the secondary will take being elevated, but I am not sure how hot that tiny xfmr will get running at it's full rating. Also, I'd be doing exactly the opposite of what is usually done. I'd be running the input stage heaters on AC, and the power tubes on DC.

Option 2: Use the above tiny transformer to isolate/elevate a linear DC filament supply (on hand). The Bridge Rectifier that is part of linear supply should provide some efficiency, although it is offset by the losses in the LM317 linear regulator. Not sure if I would "turn a profit" or not in terms of reducing load on the wimpy xfmr. Input tubes would run on well filtered, linear-regulated DC (set to 12.1V). Due to an unplanned transformer and an additional linear PS, space (and possibly heat load) under the chassis would start to be more of a problem.

Third option is to run the 12AX7 heaters (again, each tube in series, two tubes in parallel) straight from the main 12VDC power supply -- making sure that the phase inverter stage heater is the more positive end. This buys me ~6V of Heater-Cathode margin. This (and the above) has the added advantage of reducing the load on the 6V buck by 600ma.

The last option is, of course, the do nothing option. Run all filaments off the 6V 5A buck converter as originally planned, and hope that as mentioned above, the fix is the Shuguang tubes.

Please understand that while a little engineering defensiveness may be taking place, it's really that this is a non-conventional design, and thus elevating the heaters is not as straight forward as it may be with a regular BHT (Big Honkin' Transformer). I'm not arguing against the advice.

I'm leaning toward #3. Take my "free" 6V of margin, and just hope I build an amplifier and not a strobe light.

 

[peterh]

So the discussion of SMPS/Buck/Boost is in my other "Modern Power" thread.

Trust me -- I've gotten lots of comments about SMPSs.

I was hoping for comments about the rest of the design here.

Assume (right or wrong) clean DC at the set voltages.

Comments on the voltages themselves (for example screen voltage vs plate voltage) are certainly welcome.

While I do hope to be maximizing the power that can be safely gotten out of 6BQ5s (by hard-grounding the cathodes), I am not going for horsepower. More for efficiency.

I don't intend to run the plates any where near datasheet maximum.

As late coming comment; you will use one DC/DC to B+ plate and one for the screens, thus able to adjust
screen voltage independently. A good thought but with a risk: if the place DC/DC fails you will toast all
power tubes as the screens be in great risk.
I would consider another configuration : use one DC/DC for each channel, tapping the screens via a zener
diode to a proper voltage. This would half the DC/DC current and isolate the channels better.

The filement/cathode voltage of the phase splitter is high as you observed. One suggestion is to use
one ECC83 for both phasesplitters, and drive that tube from the recently shown 12V transformer leaving
the voltage-amp ecc83 ( the first in chain) powered from DC source, using one triode for each channel.

 

[peterh]

Another comment : i'm not sure you can ground the + side of your DC/DC to get a regulated (-). Also as per
the previous comment, if the (-) supply faild you will toast the tubes. Maybe a conventional catode
bias would simplify and avoid the need for a bias supply with litte loss of power. It will also
enable use of unmatched tubes if you go for individual cathode resistors.

I must repeat that i find the approach of using DC/DC supplies very interesting and breaking new grounds, in
spite of some comments elsewhere i do think that they might be an efficient way of powering tube amps.