Designing a PCB for a tube amp

[dazz]

I'm in the process of designing a pcb for a tube amp that has a JCM800 preamp and a single ended EL84 power amp. Apparently this thing has quite a lot of gain and is prone to being noisy.
I've been doing some reading on ground planes, ground loops and all that stuff and this is where I'm at (see attached pictures)

I tried to keep the power supply away from signals and the input section, and my main concern at this point is how to handle my ground connections, but any suggestion or correction will be very much appreciated. I've only designed a bunch of low signal guitar pedal pcb's, and this is my first one in Kicad, so I don't really know what I'm doing.

[andy3055]

The filament supply can introduce a lot of hum as they are AC. In the old amps, the wires are twisted so that the AC hum will get cancelled. You could probably run the filament supply on separate wires and twist them instead of putting them on the PCB. Having said that, I have noted that some guitar amps have them all on the PCB (that is the most recent I have seen a tube amp). I guess it is a matter of trace layout/design and how you can minimize the effects of AC. In any case, keep the signal traces as far as possible from the filament supply traces.

That rendering and design looks fantastic! What software did you use?

[dazz]

The filament supply can introduce a lot of hum as they are AC. In the old amps, the wires are twisted so that the AC hum will get cancelled. You could probably run the filament supply on separate wires and twist them instead of putting them on the PCB. Having said that, I have noted that some guitar amps have them all on the PCB (that is the most recent I have seen a tube amp). I guess it is a matter of trace layout/design and how you can minimize the effects of AC. In any case, keep the signal traces as far as possible from the filament supply traces.

That rendering and design looks fantastic! What software did you use?

Thanks Andy. That's Kicad's 3D viewer, really cool stuff!
I put the filament traces on the bottom layer and most of the signal traces at the top, not sure if that will be enough separation, they cross at a couple of spots. Maybe narrower filament traces would help prevent hum there? They're 1mm wide right now

[andy3055]

The filament supply can introduce a lot of hum as they are AC. In the old amps, the wires are twisted so that the AC hum will get cancelled. You could probably run the filament supply on separate wires and twist them instead of putting them on the PCB. Having said that, I have noted that some guitar amps have them all on the PCB (that is the most recent I have seen a tube amp). I guess it is a matter of trace layout/design and how you can minimize the effects of AC. In any case, keep the signal traces as far as possible from the filament supply traces.

That rendering and design looks fantastic! What software did you use?

Thanks Andy. That's Kicad's 3D viewer, really cool stuff!
I put the filament traces on the bottom layer and most of the signal traces at the top, not sure if that will be enough separation, they cross at a couple of spots. Maybe narrower filament traces would help prevent hum there? They're 1mm wide right now

Thanks for the reply. The filaments take a higher current and so, you have to keep some thickness in those traces if not they might start burning and also damaging the board. Putting them on the bottom of the board will help . Make the board as thick as possible. Since we have no experience in this area, only building it and trying will help to make a determination. Worst case, you can run wires later. Of course, you prefer to have all on the board 

[schmitt trigger]

Some people swear by DC voltage applied to filaments, specifically on high gain stages, to minimize hum.

[andy3055]

Some people swear by DC voltage applied to filaments, specifically on high gain stages, to minimize hum.

Other than the fact that you need some heavy duty diodes/bridge rectifier, I can't see why not. May be in the old days, it was too expensive to do that and they only had selenium rectifiers that were costly.

[dazz]

I have the filaments referenced to ground via resistor divider. Other than that, I don't know if there's anything else I can do to keep them from inducing noise. Hopefully I don't need to go DC for the filaments cause I would need an external power supply for that. Thanks guys 

[andy3055]

I have the filaments referenced to ground via resistor divider. Other than that, I don't know if there's anything else I can do to keep them from inducing noise. Hopefully I don't need to go DC for the filaments cause I would need an external power supply for that. Thanks guys 

Why do you need an external power supply? You already have the filament winding. The two resistor thingy with the ground tap works. Actually, you should get a wire-wound pot so that you can start at the mid point and move it one way or the other to see where you have the least hum.

All you need is a bridge that can handle the total current of the filaments.

[vk6zgo]

Some people swear by DC voltage applied to filaments, specifically on high gain stages, to minimize hum.

Other than the fact that you need some heavy duty diodes/bridge rectifier, I can't see why not. May be in the old days, it was too expensive to do that and they only had selenium rectifiers that were costly.

In tne old days, you could count on new tubes to not have heater / cathode leakage, so any injection of hum from the heater was negligible.
Sadly, these days, that is no longer the case.

On the subject of PCBs

In the early days of this technology, manufacturers seized upon it, & produced equipment with tubes on boards.
Much to their consternation, (or more correctly that of their customers), the PCBs quickly began to deteriorate  beneath the tube sockets, due to heat.

In the best case, tracks became hard to solder to, in a slightly worse case, tracks peeled, & in the worst case, delamination & charring of the board occurred, with most boards progressing through the three stages over time.


These were Phenolic boards, & FR4 is a lot better, but by no means, immune to this problem.

[floobydust]

That's some nice hard work, it's very compact.

With no negative feedback loop (from the loudspeaker back to the pre-amp) this will have much hum, noise and gain, very hard to tame. The original Marshall has push-pull with loop feedback. The EL84 would need some extra circuitry if it drives the output transformer with no NFB. I'd consider adding a radiator wire too if you like a bit of wail, something Mesa does.

A ground-pour does not cure all ills, and adds to problems with high voltages and high impedances. You must have proper spacings and better off with no ground fill on some traces/sections. Vox AC15 screwed that up and has arcing and carbon-tracking problems: https://www.eevblog.com/forum/repair/working-on-a-smoked-vox-ac15-guitar-amplifier
The via's coming from your rectifier diodes does not have decent creepage/clearance.
Roughly, 400VDC/AC is about 1.5-2mm minimum. This is just for the HV terminal block, rectifiers, B+ feeds, and plate circuit. You can have much less spacing on signals as they are low voltage, but even a few pF of PCB stray capacitance can roll off the highs.

Note your filament trace (V1, V2 pin 9) runs real close the input parts (i.e. R2) and overtop sensitive pcb traces (to C2), which is a no-no. You would have a ground guard-band around all sides of the filament trace(s) and run no sensitive signal traces crossing over top.  You get capacitive coupling of hum. DC filaments for the pre-amp tubes is the quietest; the tubes with cathode bypass capacitors it doesn't matter to them. Some product designs use flying wire only for the filaments instead of the pc board.

I don't see you grouping the high-current grounds together to keep circulating currents out of the pre-amp section. The two rectifier's anodes, C7, R20, C16 should be tied together as a "power ground". The approach is to have two star grounds, one for high current and another (quiet), then tie them together at one point. The JCM-800 used chassis points for this, Fender knew to do this even in the 1950's and point-point wiring.

I'm not sure about some part values, R2 is huge and will add noise, C3 is usually many uF. It must be a thin sound.

[TERRA Operative]

I wonder if it is worth putting some holes in tje PCB in a ring around each tube to help airflow?

[vk6zgo]

I wonder if it is worth putting some holes in tje PCB in a ring around each tube to help airflow?

That was certainly done in the old days, & helped, especially if a fan was mounted under it.
The latter wouldn't be too good on soft musical passages, though!

[floobydust]

I tried a ring of 8 of 3/16" holes and found it doesn't work, there is so little convection airflow possible through small holes, and the pcb standoffs can be too low.
Pearl tube coolers work ~50C cooler, would use on the EL84 if there's room.
Likely OP might will place this upside down in the cabinet.

[vk6zgo]

A smaller number of larger holes would be better--- that's what they did in the old days.
It still wasn't perfect, though.

[Cubdriver]

I'll second (or third or fourth) the suggestion to get the tube heater supply lines OFF of the PCB, and to run them externally using twisted pair wiring.  They draw a relatively heavy current, and are inclined to spread hum around.  Twisted pairs are somewhat 'self shielding' in that the twist distributes the radiated fields such that they tend to self-cancel.  Either put holes with pads immediately adjacent to the heater pins on the tubes and solder them there, or solder them directly to the socket pins where they come through the board.  Get them up and away from the board a bit, too.

This technique is shown below in this photo of an ST-70 clone amplifier.  Unfortuantely, I didn't take a good picture of the heater wiring specifically, but the brown twisted pair near the large silver coupling caps just to the right of center at the top of the picture shows what I'm referring to (there is also a green pair to the left, feeding the other phase splitter tube):


Nice work on the design and render!

-Pat

[dazz]

Thanks so much for you input, guys.

In tne old days, you could count on new tubes to not have heater / cathode leakage, so any injection of hum from the heater was negligible.

What a shame.You would think that, as time goes by, the manufacturing process would improve, but we live in the era of planed obsolescence. 

[Gyro]

Luckily, at least in audio stuff, the Cathode circuits are usually the lowest impedance part of the design (a lowish value cathode resistor, often bypassed). This helps to minimise the effect of heater-cathode leakage.

Biasing the heater supply to be positive relative to the cathode voltages ,say, +20-50V does help though.


P.S. You might want to consider including footprints for grid-stopper and cathode stopper series resistors for U2A. Cathode followers have a bit of a 'reputation' for poor sound quality, but this is usually down to parasitic oscillations when driving a capacitive load. You can always fit jumpers if not needed.

[dazz]

That's some nice hard work, it's very compact.

*compliments my work... proceeds to demolish it with devastating criticism* 
Thanks so much! love that

With no negative feedback loop (from the loudspeaker back to the pre-amp) this will have much hum, noise and gain, very hard to tame. The original Marshall has push-pull with loop feedback. The EL84 would need some extra circuitry if it drives the output transformer with no NFB. I'd consider adding a radiator wire too if you like a bit of wail, something Mesa does.

Yeah, I came across a message of yours in a similar thread here from 2017 where you mention that wire, but isn't it meant to do the exact opposite? isn't the wire there to push the amp into feedback oscillation instead of taming it? At any rate, it sounds like a great idea to try, and adding negative feedback too. I will google the original schematic to find out which stage does the NF plug to.

A ground-pour does not cure all ills, and adds to problems with high voltages and high impedances. You must have proper spacings and better off with no ground fill on some traces/sections. Vox AC15 screwed that up and has arcing and carbon-tracking problems: https://www.eevblog.com/forum/repair/working-on-a-smoked-vox-ac15-guitar-amplifier

Oh, I thought the issue was just stray capacitance of tracks+gnd plane rolling off high freqs. Charring sounds a lot scarier!  Unfortunately that link is broken, but I'll look that up

The via's coming from your rectifier diodes does not have decent creepage/clearance.
Roughly, 400VDC/AC is about 1.5-2mm minimum. This is just for the HV terminal block, rectifiers, B+ feeds, and plate circuit. You can have much less spacing on signals as they are low voltage, but even a few pF of PCB stray capacitance can roll off the highs.

Oh, that's something I thought I got tackled. I thought I was good with 1mm spacing for my vias since they will be coated. My PWR netclass has 1mm spacing, but that won't be enough separation between bare pads and tracks or vias, right? bummer. I'll see what I can do about that. I wonder if there's a way to set different spacings for bare pads and coated vias/tracks in Kicad

Note your filament trace (V1, V2 pin 9) runs real close the input parts (i.e. R2) and overtop sensitive pcb traces (to C2), which is a no-no. You would have a ground guard-band around all sides of the filament trace(s) and run no sensitive signal traces crossing over top.  You get capacitive coupling of hum. DC filaments for the pre-amp tubes is the quietest; the tubes with cathode bypass capacitors it doesn't matter to them. Some product designs use flying wire only for the filaments instead of the pc board.

I see what you mean. Another great piece of advice, thanks again. The first stage is good, I think, but C2 is at the input of the 2nd stage, which becomes the first in the low channel, so yeah, that's a screw up (I assumed that the signal would be large enough after the 1st stage of amplification -some 30dB- that it wouldn't matter, is that wrong?). I'm not sure there's a lot I can do about it though. No matter how I route this thing, I'm going to have some signal traces on top of the filament's, I reckon. Because it doesn't make a difference if the track that crosses the filaments is the one that goes to the grid, the cathode or the plate, right?

It's increasingly looking like I should indeed get those filament traces out of the pcb and hook them up with external wiring.

I don't see you grouping the high-current grounds together to keep circulating currents out of the pre-amp section. The two rectifier's anodes, C7, R20, C16 should be tied together as a "power ground". The approach is to have two star grounds, one for high current and another (quiet), then tie them together at one point. The JCM-800 used chassis points for this, Fender knew to do this even in the 1950's and point-point wiring.

You don't see me doing that because I had no idea about that. 
So I guess my grounding scheme is all wrong. I thought I needed to keep the power supply ground plane connected to the rest right at the star point, close to the power supply. Wouldn't it do same thing you suggest if I added another keepout polygon between the output tube and the preamp section?

[dazz]

Regarding heat management and adding holes around the power tube, I read somewhere during my research that ceramic sockets help a lot. Here's hoping that's true, cause I don't have enough room for holes.

EDIT:

I tried a ring of 8 of 3/16" holes and found it doesn't work, there is so little convection airflow possible through small holes, and the pcb standoffs can be too low.
Pearl tube coolers work ~50C cooler, would use on the EL84 if there's room.
Likely OP might will place this upside down in the cabinet.

Oh!, that's pretty neat! I'll see if I can find them here

[exe]

I'm not sure the ham is induced by the traces. The frequency is only 50/60Hz, so I'd be surprised that a 20-30cm trace would be an effective antenna. I'd expect ham to come from the environment, or somehow else gets coupled into the circuit, but not directly radiated by traces/wires.

[dazz]

Luckily, at least in audio stuff, the Cathode circuits are usually the lowest impedance part of the design (a lowish value cathode resistor, often bypassed). This helps to minimise the effect of heater-cathode leakage.

Biasing the heater supply to be positive relative to the cathode voltages ,say, +20-50V does help though.


P.S. You might want to consider including footprints for grid-stopper and cathode stopper series resistors for U2A. Cathode followers have a bit of a 'reputation' for poor sound quality, but this is usually down to parasitic oscillations when driving a capacitive load. You can always fit jumpers if not needed.

Thanks Gyro, I'll need to look up "cathode stopper", that's a new one on me

[Gyro]

You're welcome. Basically it is a low value series resistor that helps isolate the cathode (well the grid-cathode voltage) from the capacitive load. The ECC83 is a fairly high gain tube which is being used as a unity gain buffer in a cathode follower so stability can become an issue. You will often see similar things in transistor emitter followers and opamp outputs where unity gain into a capacitive load can be an issue. go for something like 47 - 100 ohms, that doesn't make any practical difference to the circuit impedances.

The golden rule for stopper resistors is to put them as close to the tube as sensibly possible.

[dazz]

You're welcome. Basically it is a low value series resistor that helps isolate the cathode (well the grid-cathode voltage) from the capacitive load. The ECC83 is a fairly high gain tube which is being used as a unity gain buffer in a cathode follower so stability can become an issue. You will often see similar things in transistor emitter followers and opamp outputs where unity gain into a capacitive load can be an issue. go for something like 47 - 100 ohms, that doesn't make any practical difference to the circuit impedances.

The golden rule for stopper resistors is to put them as close to the tube as sensibly possible.

Added the grid stopper, but I still don't understand what a cathode stopper is. I already have a cathode resistor in R17, would it be a resistor between the grid and the cathode?

[Gyro]

No, just a series resistor on the cathode, just like the one on the grid (doesn't matter whether you put it before the junction with R17 or after).

P.S. I'm really talking belt and braces precautions here.

[dazz]

No, just a series resistor on the cathode, just like the one on the grid (doesn't matter whether you put it before the junction with R17 or after).

P.S. I'm really talking belt and braces precautions here.

Ah, I see what you mean now. Sorry, I'm a bit slow sometimes

ETA: And of course, the capacitive load you were referring to is the tone stack. It's starting to sink! finally!