Knocking back B+ for 12AX7 tube project.

[digi2t]

Rob,

Thanks a ton for breaking that down for me. As always, I'm in awe of how some of you guys can just breeze through some of this stuff, but more so when you can also dumb it down for us hackers.

At this point, I do have a toroidal transfo on the way, and I'm planning on giving it a go. It seems to be a tested go to, so far be it from me to try and reinvent the wheel. I would like to test the EI transfos as well though, but I think that will require long term testing to carefully monitor how they performs (with CO2 extinguisher at the ready maybe? ).

Just to touch back on the subject of the cap on the output for the EI transfo; is there a rough rule of thumb insofar as sizing the cap is concerned, or is it a "try and test" scenario? Would it have to be one of those so called "safety caps" (X1), or will any film cap of adequate rating do?

[Rob Strand]

It seems to be a tested go to, so far be it from me to try and reinvent the wheel. I would like to test the EI transfos as well though, but I think that will require long term testing to carefully monitor how they performs (with CO2 extinguisher at the ready maybe? ).

It's more of a slow death.  In bad cases it could fry if left for 1 to 8 hrs.   In other cases it will show up as a lot of in-field failures over time.   Some of those Hughes-Kettner pedals have a history of just riding this point.

Just to touch back on the subject of the cap on the output for the EI transfo; is there a rough rule of thumb insofar as sizing the cap is concerned, or is it a "try and test" scenario? Would it have to be one of those so called "safety caps" (X1), or will any film cap of adequate rating do?

It's very much a "try and test" for each specific transformer.   A simple algorithm is to measure the input current to the transformer, then find the cap that has the least current.  No need for small steps, doubling or halving the cap value is good enough.   The value of the cap may need to be backed-off when you put the rectifier and filter on the output so the tuning is best done in the final build.

It's best to use a smaller resonating cap than a larger one.  If you have to pick between two values pick a smaller cap.

Even if you can shave 20% off the input current the winding will run cooler by a factor of 1/1.2^2 = 0.69, which could be enough to save it.

I pulled out some old measurements of two EI35x10 core transformers:

Transformer 1:
230V winding, 50Hz, 1.8VA, HV winding resistance 1340 ohm, 
Optimum cap on HV winding 220nF,   LV Current with cap is 0.24 times LV current without cap.
With 100nF,  LV Current with cap is 0.57 times LV current without cap.

Transformer 2:
230V winding, 50Hz, 1.35VA, HV winding resistance 3140 ohm
Optimum cap on HV winding 50nF,   LV Current with cap is 0.68 times LV current without cap.
With 22nF,  LV Current with cap is 0.81 times LV current without cap.
*** With 100nF LV winding current is *higher* than no having the cap at all.   That emphasizes the need to go for smaller rather than larger caps.

To translate the cap values to a 115V winding you need to multiply the capacitance by 4 (ie. (230/115)^2).
That gives 400nF to 820nF for transformer 1 and 88nF to 200nF for transformer 2.   50Hz to 60Hz would mean the cap would get smaller by a factor (60/50)^2 = 1.44.

Transformer 1 is a real stinker and pulls a heap of LV current in reverse.   The low HV winding resistance is a sign they used less turns of thicker wire so the core is pushed harder and that makes the no-load current go up.    However, the low resistance is probably why it has more reduction in current with the cap.

Transformer 2 is completely different design to transformer 1.   The HV windings resistance is relatively high.  This transformer would probably work fine without the added cap.    The high winding resistance is probably why there is less reduction in current with the cap.   Still, the reduction isn't bad.

I didn't do any tests with the cap on the LV side.  The caps will be quite large.   I'd guess this method might help Transformer 2 a bit more.

Something worth mentioning is how consistent the amount of current reduction is across transformers of the same type with the same cap value.   I never got around to checking that.   I have a number of 2VA transformers with the same specs but they are different designs over a 40 year period.  The thing to check is if the 'best' cap on average causes the current to go up on the worst transformer.  That would give an idea how much smaller we should make the cap compared to the best one.


FWIW,  transformer 2 has more resistance because it has more turns of thinner wire.  More turns means a higher inductance.   The resonant idea tunes the resonant frequency to the mains frequency so we can sort of see why transformer 2 requires a smaller cap;  C = 1 / ( (2*pi*f)^2 * L) .   If you measure the transformer inductance with an LCR meter it's likely the estimated capacitance will be way off - I don't recommend this method.

[digi2t]

So, an update...

Been noodling with this project for a couple of months now. Finally settled on a potted toroidal. Using a 12vac power supply, I found that the 115v/115v - 15v/15v works really well. I get about 185vac for the B+, and no problem supplying an adequate 12vac for the 12v and 5v regulators.

Since there were errors on the PCB, I ended up having to hack it to test with, which meant mounting the transfo in a vise next to the board, with jumper wires between the two. No doubt, the jumpers induced a bunch of noise into the works, specifically from the HV side. Moving my hand close to, and away from the wires, you can hear the buzzing change. I'm assuming that with the transfo mounted to the PCB, and the bridge in tight proximity to keep the traces uber-short, should reduce the noise quite a bit.

My question though is regarding ground plane(s) under the transformer. I'm using a 70003K transformer, and I'm wondering what the protocol is regarding ground planes in this case. Yes, or no? Will it help with noise reduction, or create an induction issue with the transformer?

[anotherjim]

Toroids are often clamped down on a metal chassis. I don't see a PCB ground pour being any different. The only thing to avoid is creating a short circuit turn through the toroid with a centre bolt fixing, which is probably impossible with the one you have?

[digi2t]

Toroids are often clamped down on a metal chassis. I don't see a PCB ground pour being any different. The only thing to avoid is creating a short circuit turn through the toroid with a centre bolt fixing, which is probably impossible with the one you have?

I'm using the potted version, like this;



So no issues with shorting, since it's only a 4mm screw into the plastic that secures it. I was only concerned with any potential induction issues.

[GibsonGM]

FWIW, I've used a power Mosfet VVR to reduce B+ in my 18 Watt, worked great til I removed it for some reason! (probably because the magic of that amp is hitting the EL84s hard and power scaling didn't do that very well).  Very simple solution, if applicable to your situation; schematics readily available via search engine...

[digi2t]

That's good to keep in mind. In this case, I'm also trying to keep the parts count to a minimum, or at the very least identical to what's used in the original. Swapping out the transformer for a higher voltage primary does the job. B+ is now sitting at 185vdc, which is well within the caps voltage capacity, and only about 20vdc higher than the original circuit.

Same project, different topic...

PCB traces for AC power. What's the best routing tactic for routing AC traces for low noise? I'm thinking of AC heater wiring in a tube amp, and how they're twisted together to cancel out noise. On a PCB, what's best practice; parallel traces on the same side of the board (at what spacing?), or one directly above the other on opposite sides of the board?

[merlinb]

On a PCB, what's best practice; parallel traces on the same side of the board (at what spacing?), or one directly above the other on opposite sides of the board?

In most cases, parallel traces, same side, as close together as possible assuming low voltage. Obviously you should observe creepage rules if you're running high voltage.
If it's a long run you can 'stitch' a pair through the board every few inches to create a quasi-twisted arrangement.
If you need a LOT of copper for high current you can run two parallel pairs on opposite sides of the board, directly on top of one another and each with opposite polarity to the pair on the other side of the board. But I doubt you're talking about that sort of current level.

[digi2t]

OK, thanks Merlin.

Funny you mentioned it, but my mind was wandering last night, and the "stitched" trace idea actually came to mind, but I didn't know if that was even a thing. Interesting!

Anyway, you assume correctly, we're not talking big currents here. I'm just trying to take whatever measures I can to make sure this is quiet.

[puretube]

You mentioned 12V & 5V regulators: so go DC for the filaments, too!!!
Reread reply#13, and ask yourself or others whether a thin foil of copper (PCB) shields magnetic fields.

[digi2t]

You mentioned 12V & 5V regulators: so go DC for the filaments, too!!!
Reread reply#13, and ask yourself or others whether a thin foil of copper (PCB) shields magnetic fields.

Indeed, the filaments are to be powered by 12vdc.